An increasing number of applications are driven by large and complex neural networks trained on diverse sets of accelerators. This process is facilitated by ML compilers that map high-level computational graphs to low-level, device-specific executables. In doing so, ML compilers need to solve many optimization problems, including graph rewriting, assignment of operations on devices, operation fusion, layout and tiling of tensors, and scheduling. For example, in a device placement problem, the compiler needs to determine the mapping between operations in the computational graph to the target physical devices so that an objective function, such as training step time, can be minimized. The placement performance is determined by a mixture of intricate factors, including inter-device network bandwidth, peak device memory, co-location constraints, etc., making it challenging for heuristics or search-based algorithms, which typically settle for fast, but sub-optimal, solutions. Furthermore, heuristics are hard to develop and maintain, especially as newer model architectures emerge.
Recent attempts at using learning-based approaches have demonstrated promising results, but they have a number of limitations that make them infeasible to be deployed in practice. Firstly, these approaches do not easily generalize to unseen graphs, especially those arising from newer model architectures, and second, they have poor sample efficiency, leading to high resource consumption during training. Finally, they are only able to solve a single optimization task, and consequently, do not capture the dependencies across the tightly coupled optimization problems in the compilation stack.
In “Transferable Graph Optimizers for ML Compilers”, recently published as an oral paper at NeurIPS 2020, we propose an end-to-end, transferable deep reinforcement learning method for computational graph optimization (GO) that overcomes all of the above limitations. We demonstrate 33%-60% speedup on three graph optimization tasks compared to TensorFlow default optimization. On a diverse set of representative graphs consisting of up to 80,000 nodes, including Inception-v3, Transformer-XL, and WaveNet, GO achieves an average 21% improvement over expert optimization and an 18% improvement over the prior state of the art with 15x faster convergence.
Graph Optimization Problems in ML Compilers There are three coupled optimization tasks that frequently arise in ML compilers, which we formulate as decision problems that can be solved using a learned policy. The decision problems for each of the tasks can be reframed as making a decision for each node in the computational graph.
The first optimization task is device placement, where the goal is to determine how best to assign the nodes of the graph to the physical devices on which it runs such that the end-to-end run time is minimized.
The second optimization task is operation scheduling. An operation in a computational graph is ready to run when its incoming tensors are present in the device memory. A frequently used scheduling strategy is to maintain a ready queue of operations for each device and schedule operations in first-in-first-out order. However, this scheduling strategy does not take into account the downstream operations placed on other devices that might be blocked by an operation, and often leads to schedules with underutilized devices. To find schedules that can keep track of such cross-device dependencies, our approach uses a priority-based scheduling algorithm that schedules operations in the ready queue based on the priority of each. Similar to device placement, operation scheduling can then be formulated as the problem of learning a policy that assigns a priority for each node in the graph to maximize a reward based on run time.
The third optimization task is operation fusion. For brevity we omit a detailed discussion of this problem here, and instead just note that similar to priority-based scheduling, operation fusion can also use a priority-based algorithm to decide which nodes to fuse. The goal of the policy network in this case is again to assign a priority for each node in the graph.
Finally, it is important to recognize that the decisions taken in each of the three optimization problems can affect the optimal decision for the other problems. For example, placing two nodes on two different devices effectively disables fusion and introduces a communication delay that can influence scheduling.
RL Policy Network Architecture Our research presents GO, a deep RL framework that can be adapted to solve each of the aforementioned optimization problems — both individually as well as jointly. There are three key aspects of the proposed architecture:
First, we use graph neural networks (specifically GraphSAGE) to capture the topological information encoded in the computational graph. The inductive network of GraphSAGE leverages node attribute information to generalize to previously unseen graphs, which enables decision making for unseen data without incurring significant cost on training.
Second, computational graphs for many models often contain more than 10k nodes. Solving the optimization problems effectively over such large scales requires that the network is able to capture long-range dependencies between nodes. GO’s architecture includes a scalable attention network that uses segment-level recurrence to capture such long-range node dependencies.
Third, ML compilers need to solve optimization problems over a wide variety of graphs from different application domains. A naive strategy of training a shared policy network with heterogeneous graphs is unlikely to capture the idiosyncrasies of a particular class of graphs. To overcome this, GO uses a feature modulation mechanism that allows the network to specialize for specific graph types without increasing the number of parameters.
To jointly solve multiple dependent optimization tasks, GO has the ability to add additional recurrent attention layers for each task with parameters shared across different tasks. The recurrent attention layers with residual connections of actions enables tracking inter-task dependencies.
Results Next, we present evaluation results on a single-task speedup on a device placement task based on real-hardware measurements, generalization to unseen graphs with different GO variants, and multi-task performance jointly optimizing operations fusion, device placement, and scheduling.
Speedup: To evaluate the performance of this architecture, we apply GO to a device placement problem based on real-hardware evaluation, where we first train the model separately on each of our workloads. This approach, called GO-one, consistently outperforms expert manual placement (HP), TensorFlow METIS placement, and Hierarchical Device Placement (HDP) — the current state-of-the-art reinforcement learning-based device placement. Importantly, with the efficient end-to-end single-shot placement, GO-one has a 15x speedup in convergence time of the placement network over HDP.
Our empirical results show that GO-one consistently outperforms expert placement, TensorFlow METIS placement, and hierarchical device placement (HDP). Because GO is designed in a way to scale up to extremely large graphs consisting of over 80,000 nodes like an 8-layer Google Neural Machine Translation (GNMT) model, it outperforms previous approaches, including HDP, REGAL, and Placeto. GO achieves optimized graph runtimes for large graphs like GNMT that are 21.7% and 36.5% faster than HP and HDP, respectively. Overall, GO-one achieves on average 20.5% and 18.2% run time reduction across a diverse set of 14 graphs, compared to HP and HDP respectively. Importantly, with the efficient end-to-end single-shot placement, GO-one has a 15x speedup in convergence time of the placement network over HDP.
Generalization: GO generalizes to unseen graphs using offline pre-training followed by fine-tuning on the unseen graphs. During pre-training, we train GO on heterogeneous subsets of graphs from the training set. We train GO for 1000 steps on each such batch of graphs before switching to the next. This pretrained model is then fine-tuned (GO-generalization+finetune) on hold-out graphs for fewer than 50 steps, which typically takes less than one minute. GO-generalization+finetune for hold-out graphs outperforms both expert placement and HDP consistently on all datasets, and on average matches GO-one.
We also run inference directly on just the pre-trained model without any fine-tuning for the target hold-out graphs, and name this GO-generalization-zeroshot. The performance of this untuned model is only marginally worse than GO-generalization+finetune, while being slightly better than expert placement and HDP. This indicates that both graph embedding and the learned policies transfer efficiently, allowing the model to generalize to the unseen data.
Co-optimizing placement, scheduling, and fusion (pl+sch+fu): Optimizing simultaneously for placement, scheduling and fusion provides 30%-73% speedup compared to the single-gpu unoptimized case and 33%-60% speedup compared to TensorFlow default placement, scheduling, and fusion. Comparing to optimizing each tasks individually, multi-task GO (pl+sch+fu) outperforms single-task GO (p | sch | fu) — optimizing all tasks, one at a time — by an average of 7.8%. Furthermore, for all workloads, co-optimizing all three tasks offers faster run time than optimizing any two of them and using the default policy for the third.
Conclusion The increasing complexity and diversity of hardware accelerators has made the development of robust and adaptable ML frameworks onerous and time-consuming, often requiring multiple years of effort from hundreds of engineers. In this article, we demonstrated that many of the optimization problems in such frameworks can be solved efficiently and optimally using a carefully designed learned approach.
Acknowledgements This is joint work with Daniel Wong, Amirali Abdolrashidi, Peter Ma, Qiumin Xu, Hanxiao Liu, Mangpo Phitchaya Phothilimthana, Shen Wang, Anna Goldie, Azalia Mirhoseini, and James Laudon.
Machine learning-based language models trained to predict the next word in a sentence have become increasingly capable, common, and useful, leading to groundbreaking improvements in applications like question-answering, translation, and more. But as language models continue to advance, new and unexpected risks can be exposed, requiring the research community to proactively work to develop new ways to mitigate potential problems.
One such risk is the potential for models to leak details from the data on which they’re trained. While this may be a concern for all large language models, additional issues may arise if a model trained on private data were to be made publicly available. Because these datasets can be large (hundreds of gigabytes) and pull from a range of sources, they can sometimes contain sensitive data, including personally identifiable information (PII) — names, phone numbers, addresses, etc., even if trained on public data. This raises the possibility that a model trained using such data could reflect some of these private details in its output. It is therefore important to identify and minimize the risks of such leaks, and to develop strategies to address the issue for future models.
In “Extracting Training Data from Large Language Models”, a collaboration with OpenAI, Apple, Stanford, Berkeley, and Northeastern University, we demonstrate that, given only the ability to query a pre-trained language model, it is possible to extract specific pieces of training data that the model has memorized. As such, training data extraction attacks are realistic threats on state-of-the-art large language models. This research represents an early, critical step intended to inform researchers about this class of vulnerabilities, so that they may take steps to mitigate these weaknesses.
Ethics of Language Model Attacks A training data extraction attack has the greatest potential for harm when applied to a model that is available to the public, but for which the dataset used in training is not. However, since conducting this research on such a dataset could have harmful consequences, we instead mount a proof of concept training data extraction attack on GPT-2, a large, publicly available language model developed by OpenAI, that was trained using only public data. While this work focuses on GPT-2 specifically, the results apply to understanding what privacy threats are possible on large language models generally.
As with other privacy- and security-related research, it is important to consider the ethics of such attacks before actually performing them. To minimize the potential risk of this work, the training data extraction attack in this work was developed using publicly available data. Furthermore, the GPT-2 model itself was made public by OpenAI in 2019, and the training data used to train GPT-2 was collected from the public internet, and is available for download by anyone who follows the data collection process documented in the GPT-2 paper.
Additionally, in accordance with responsible computer security disclosure norms, we followed up with individuals whose PII was extracted, and secured their permission before including references to this data in publication. Further, in all publications of this work, we have redacted any personally identifying information that may identify individuals. We have also worked closely with OpenAI in the analysis of GPT-2.
The Training Data Extraction Attack By design, language models make it very easy to generate a large amount of output data. By seeding the model with random short phrases, the model can generate millions of continuations, i.e., probable phrases that complete the sentence. Most of the time, these continuations will be benign strings of sensible text. For example, when asked to predict the continuation of the string “Mary had a little…”, a language model will have high confidence that the next token is the word “lamb”. However, if one particular training document happened to repeat the string “Mary had a little wombat” many times, the model might predict that phrase instead.
The goal of a training data extraction attack is then to sift through the millions of output sequences from the language model and predict which text is memorized. To accomplish this, our approach leverages the fact that models tend to be more confident on results captured directly from their training data. These membership inference attacks enable us to predict if a result was used in the training data by checking the confidence of the model on a particular sequence.
The main technical contribution of this work is the development of a method for inferring membership with high accuracy along with techniques for sampling from models in a way that encourages the output of memorized content. We tested a number of different sampling strategies, the most successful of which generates text conditioned on a wide variety of input phrases. We then compare the output of two different language models. When one model has high confidence in a sequence, but the other (equally accurate) model has low confidence in a sequence, it's likely that the first model has memorized the data.
Results Out of 1800 candidate sequences from the GPT-2 language model, we extracted over 600 that were memorized from the public training data, with the total number limited by the need for manual verification. The memorized examples cover a wide range of content, including news headlines, log messages, JavaScript code, PII, and more. Many of these examples are memorized even though they appear infrequently in the training dataset. For example, for many samples of PII we extract are found in only a single document in the dataset. However, in most of these cases, the originating document contains multiple instances of the PII, and as a result, the model still learns it as high likelihood text.
Finally, we also find that the larger the language model, the more easily it memorizes training data. For example, in one experiment we find that the 1.5 billion parameter GPT-2 XL model memorizes 10 times more information than the 124 million parameter GPT-2 Small model. Given that the research community has already trained models 10 to 100 times larger, this means that as time goes by, more work will be required to monitor and mitigate this problem in increasingly large language models.
Lessons While we demonstrate these attacks on GPT-2 specifically, they show potential flaws in all large generative language models. The fact that these attacks are possible has important consequences for the future of machine learning research using these types of models.
Fortunately, there are several ways to mitigate this issue. The most straightforward solution is to ensure that models do not train on any potentially problematic data. But this can be difficult to do in practice.
The use of differential privacy, which allows training on a dataset without revealing any details of individual training examples, is one of the most principled techniques to train machine learning models with privacy. In TensorFlow, this can be achieved with the use of the tensorflow/privacy module (or similar for PyTorch or JAX) that is a drop-in replacement for existing optimizers. Even this can have limitations and won’t prevent memorization of content that is repeated often enough. If this is not possible, we recommend at least measuring how much memorization occurs so appropriate action can be taken.
Language models continue to demonstrate great utility and flexibility—yet, like all innovations, they can also pose risks. Developing them responsibly means proactively identifying those risks and developing ways to mitigate them. We hope that this effort to highlight current weaknesses in large language modeling will raise awareness of this challenge in the broader machine learning community and motivate researchers to continue to develop effective techniques to train models with reduced memorization.
Acknowledgements This work was performed jointly with Florian Tramer, Eric Wallace, Matthew Jagielski, Ariel Herbert-Voss, Katherine Lee, Adam Roberts, Tom Brown, Dawn Song, Ulfar Erlingsson, Alina Oprea, and Colin Raffel.
Professional portrait photographers are able to create compelling photographs by using specialized equipment, such as off-camera flashes and reflectors, and expert knowledge to capture just the right illumination of their subjects. In order to allow users to better emulate professional-looking portraits, we recently released Portrait Light, a new post-capture feature for the Pixel Camera and Google Photos apps that adds a simulated directional light source to portraits, with the directionality and intensity set to complement the lighting from the original photograph.
In the Pixel Camera on Pixel 4, Pixel 4a, Pixel 4a (5G), and Pixel 5, Portrait Light is automatically applied post-capture to images in the default mode and to Night Sight photos that include people — just one person or even a small group. In Portrait Mode photographs, Portrait Light provides more dramatic lighting to accompany the shallow depth-of-field effect already applied, resulting in a studio-quality look. But because lighting can be a personal choice, Pixel users who shoot in Portrait Mode can manually re-position and adjust the brightness of the applied lighting within Google Photos to match their preference. For those running Google Photos on Pixel 2 or newer, this relighting capability is also available for many pre-existing portrait photographs.
Today we present the technology behind Portrait Light. Inspired by the off-camera lights used by portrait photographers, Portrait Light models a repositionable light source that can be added into the scene, with the initial lighting direction and intensity automatically selected to complement the existing lighting in the photo. We accomplish this by leveraging novel machine learning models, each trained using a diverse dataset of photographs captured in the Light Stage computational illumination system. These models enabled two new algorithmic capabilities:
These innovations enable Portrait Light to help create attractive lighting at any moment for every portrait — all on your mobile device.
Automatic Light Placement Photographers usually rely on perceptual cues when deciding how to augment environmental illumination with off-camera light sources. They assess the intensity and directionality of the light falling on the face, and also adjust their subject’s head pose to complement it. To inform Portrait Light’s automatic light placement, we developed computational equivalents to these two perceptual signals.
First, we trained a novel machine learning model to estimate a high dynamic range, omnidirectional illumination profile for a scene based on an input portrait. This new lighting estimation model infers the direction, relative intensity, and color of all light sources in the scene coming from all directions, considering the face as a light probe. We also estimate the head pose of the portrait’s subject using MediaPipe Face Mesh.
Using these clues, we determine the direction from which the synthetic lighting should originate. In studio portrait photography, the main off-camera light source, or key light, is placed about 30° above the eyeline and between 30° and 60° off the camera axis, when looking overhead at the scene. We follow this guideline for a classic portrait look, enhancing any pre-existing lighting directionality in the scene while targeting a balanced, subtle key-to-fill lighting ratio of about 2:1.
Data-Driven Portrait Relighting Given a desired lighting direction and portrait, we next trained a new machine learning model to add the illumination from a directional light source to the original photograph. Training the model required millions of pairs of portraits both with and without extra light. Photographing such a dataset in normal settings would have been impossible because it requires near-perfect registration of portraits captured across different lighting conditions.
Instead, we generated training data by photographing seventy different people using the Light Stage computational illumination system. This spherical lighting rig includes 64 cameras with different viewpoints and 331 individually-programmable LED light sources. We photographed each individual illuminated one-light-at-a-time (OLAT) by each light, which generates their reflectance field — or their appearance as illuminated by the discrete sections of the spherical environment. The reflectance field encodes the unique color and light-reflecting properties of the subject’s skin, hair, and clothing — how shiny or dull each material appears. Due to the superposition principle for light, these OLAT images can then be linearly added together to render realistic images of the subject as they would appear in any image-based lighting environment, with complex light transport phenomena like subsurface scattering correctly represented.
Using the Light Stage, we photographed many individuals with different face shapes, genders, skin tones, hairstyles, and clothing/accessories. For each person, we generated synthetic portraits in many different lighting environments, both with and without the added directional light, rendering millions of pairs of images. This dataset encouraged model performance across diverse lighting environments and individuals.
Learning Detail-Preserving Relighting Using the Quotient Image Rather than trying to directly predict the output relit image, we trained the relighting model to output a low-resolution quotient image, i.e., a per-pixel multiplier that when upsampled can be applied to the original input image to produce the desired output image with the contribution of the extra light source added. This technique is computationally efficient and encourages only low-frequency lighting changes, without impacting high-frequency image details, which are directly transferred from the input to maintain image quality.
Supervising Relighting with Geometry Estimation When photographers add an extra light source into a scene, its orientation relative to the subject’s facial geometry determines how much brighter each part of the face appears. To model the optical behavior of light sources reflecting off relatively matte surfaces, we first trained a machine learning model to estimate surface normals given the input photograph, and then applied Lambert’s law to compute a “light visibility map” for the desired lighting direction. We provided this light visibility map as input to the quotient image predictor, ensuring that the model is trained using physics-based insights.
We optimized the full pipeline to run at interactive frame-rates on mobile devices, with total model size under 10 MB. Here are a few examples of Portrait Light in action.
Getting the Most Out of Portrait Light You can try Portrait Light in the Pixel Camera and change the light position and brightness to your liking in Google Photos. For those who use Dual Exposure Controls, Portrait Light can be applied post-capture for additional creative flexibility to find just the right balance between light and shadow. On existing images from your Google Photos library, try it where faces are slightly underexposed, where Portrait Light can illuminate and highlight your subject. It will especially benefit images with a single individual posed directly at the camera.
We see Portrait Light as the first step on the journey towards creative post-capture lighting controls for mobile cameras, powered by machine learning.
Acknowledgements Portrait Light is the result of a collaboration between Google Research, Google Daydream, Pixel, and Google Photos teams. Key contributors include: Yun-Ta Tsai, Rohit Pandey, Sean Fanello, Chloe LeGendre, Michael Milne, Ryan Geiss, Sam Hasinoff, Dillon Sharlet, Christoph Rhemann, Peter Denny, Kaiwen Guo, Philip Davidson, Jonathan Taylor, Mingsong Dou, Pavel Pidlypenskyi, Peter Lincoln, Jay Busch, Matt Whalen, Jason Dourgarian, Geoff Harvey, Cynthia Herrera, Sergio Orts Escolano, Paul Debevec, Jonathan Barron, Sofien Bouaziz, Clement Ng, Rachit Gupta, Jesse Evans, Ryan Campbell, Sonya Mollinger, Emily To, Yichang Shih, Jana Ehmann, Wan-Chun Alex Ma, Christina Tong, Tim Smith, Tim Ruddick, Bill Strathearn, Jose Lima, Chia-Kai Liang, David Salesin, Shahram Izadi, Navin Sarma, Nisha Masharani, Zachary Senzer.
1 Work conducted while at Google. ↩
Real-time, simultaneous perception of human pose, face landmarks and hand tracking on mobile devices can enable a variety of impactful applications, such as fitness and sport analysis, gesture control and sign language recognition, augmented reality effects and more. MediaPipe, an open-source framework designed specifically for complex perception pipelines leveraging accelerated inference (e.g., GPU or CPU), already offers fast and accurate, yet separate, solutions for these tasks. Combining them all in real-time into a semantically consistent end-to-end solution is a uniquely difficult problem requiring simultaneous inference of multiple, dependent neural networks.
Today, we are excited to announce MediaPipe Holistic, a solution to this challenge that provides a novel state-of-the-art human pose topology that unlocks novel use cases. MediaPipe Holistic consists of a new pipeline with optimized pose, face and hand components that each run in real-time, with minimum memory transfer between their inference backends, and added support for interchangeability of the three components, depending on the quality/speed tradeoffs. When including all three components, MediaPipe Holistic provides a unified topology for a groundbreaking 540+ keypoints (33 pose, 21 per-hand and 468 facial landmarks) and achieves near real-time performance on mobile devices. MediaPipe Holistic is being released as part of MediaPipe and is available on-device for mobile (Android, iOS) and desktop. We are also introducing MediaPipe’s new ready-to-use APIs for research (Python) and web (JavaScript) to ease access to the technology.
Pipeline and Quality The MediaPipe Holistic pipeline integrates separate models for pose, face and hand components, each of which are optimized for their particular domain. However, because of their different specializations, the input to one component is not well-suited for the others. The pose estimation model, for example, takes a lower, fixed resolution video frame (256x256) as input. But if one were to crop the hand and face regions from that image to pass to their respective models, the image resolution would be too low for accurate articulation. Therefore, we designed MediaPipe Holistic as a multi-stage pipeline, which treats the different regions using a region appropriate image resolution.
First, MediaPipe Holistic estimates the human pose with BlazePose’s pose detector and subsequent keypoint model. Then, using the inferred pose key points, it derives three regions of interest (ROI) crops for each hand (2x) and the face, and employs a re-crop model to improve the ROI (details below). The pipeline then crops the full-resolution input frame to these ROIs and applies task-specific face and hand models to estimate their corresponding keypoints. Finally, all key points are merged with those of the pose model to yield the full 540+ keypoints.
To streamline the identification of ROIs, a tracking approach similar to the one used for the standalone face and hand pipelines is utilized. This approach assumes that the object doesn't move significantly between frames, using an estimation from the previous frame as a guide to the object region in the current one. However, during fast movements, the tracker can lose the target, which requires the detector to re-localize it in the image. MediaPipe Holistic uses pose prediction (on every frame) as an additional ROI prior to reduce the response time of the pipeline when reacting to fast movements. This also enables the model to retain semantic consistency across the body and its parts by preventing a mixup between left and right hands or body parts of one person in the frame with another.
In addition, the resolution of the input frame to the pose model is low enough that the resulting ROIs for face and hands are still too inaccurate to guide the re-cropping of those regions, which require a precise input crop to remain lightweight. To close this accuracy gap we use lightweight face and hand re-crop models that play the role of spatial transformers and cost only ~10% of the corresponding model's inference time.
Performance MediaPipe Holistic requires coordination between up to 8 models per frame — 1 pose detector, 1 pose landmark model, 3 re-crop models and 3 keypoint models for hands and face. While building this solution, we optimized not only machine learning models, but also pre- and post-processing algorithms (e.g., affine transformations), which take significant time on most devices due to pipeline complexity. In this case, moving all the pre-processing computations to GPU resulted in ~1.5 times overall pipeline speedup depending on the device. As a result, MediaPipe Holistic runs in near real-time performance even on mid-tier devices and in the browser.
The multi-stage nature of the pipeline provides two more performance benefits. As models are mostly independent, they can be replaced with lighter or heavier versions (or turned off completely) depending on the performance and accuracy requirements. Also, once pose is inferred, one knows precisely whether hands and face are within the frame bounds, allowing the pipeline to skip inference on those body parts.
Applications MediaPipe Holistic, with its 540+ key points, aims to enable a holistic, simultaneous perception of body language, gesture and facial expressions. Its blended approach enables remote gesture interfaces, as well as full-body AR, sports analytics, and sign language recognition. To demonstrate the quality and performance of the MediaPipe Holistic, we built a simple remote control interface that runs locally in the browser and enables a compelling user interaction, no mouse or keyboard required. The user can manipulate objects on the screen, type on a virtual keyboard while sitting on the sofa, and point to or touch specific face regions (e.g., mute or turn off the camera). Underneath it relies on accurate hand detection with subsequent gesture recognition mapped to a "trackpad" space anchored to the user’s shoulder, enabling remote control from up to 4 meters.
This technique for gesture control can unlock various novel use-cases when other human-computer interaction modalities are not convenient. Try it out in our web demo and prototype your own ideas with it.
MediaPipe for Research and Web To accelerate ML research as well as its adoption in the web developer community, MediaPipe now offers ready-to-use, yet customizable ML solutions in Python and in JavaScript. We are starting with those in our previous publications: Face Mesh, Hands and Pose, including MediaPipe Holistic, with many more to come. Try them directly in the web browser: for Python using the notebooks in MediaPipe on Google Colab, and for JavaScript with your own webcam input in MediaPipe on CodePen!
Conclusion We hope the release of MediaPipe Holistic will inspire the research and development community members to build new unique applications. We anticipate that these pipelines will open up avenues for future research into challenging domains, such as sign-language recognition, touchless control interfaces, or other complex use cases. We are looking forward to seeing what you can build with it!
Acknowledgments Special thanks to all our team members who worked on the tech with us: Fan Zhang, Gregory Karpiak, Kanstantsin Sokal, Juhyun Lee, Hadon Nash, Chuo-Ling Chang, Jiuqiang Tang, Nikolay Chirkov, Camillo Lugaresi, George Sung, Michael Hays, Tyler Mullen, Chris McClanahan, Ekaterina Ignasheva, Marat Dukhan, Artsiom Ablavatski, Yury Kartynnik, Karthik Raveendran, Andrei Vakunov, Andrei Tkachenka, Suril Shah, Buck Bourdon, Ming Guang Yong, Esha Uboweja, Siarhei Kazakou, Andrei Kulik, Matsvei Zhdanovich, and Matthias Grundmann.
This week marks the beginning of the 34th annual Conference on Neural Information Processing Systems (NeurIPS 2020), the biggest machine learning conference of the year. Held virtually for the first time, this conference includes invited talks, demonstrations and presentations of some of the latest in machine learning research. As a Platinum Sponsor of NeurIPS 2020, Google will have a strong presence with more than 180 accepted papers, additionally contributing to and learning from the broader academic research community via talks, posters, workshops and tutorials.
If you are registered for NeurIPS 2020, we hope you’ll visit our virtual booth and chat with our researchers about the projects and opportunities at Google that go into solving the world's most challenging research problems, and to see demonstrations of some of the exciting research we pursue, such as Transformers for image recognition, Tone Transfer, large-scale distributed RL, recreating historical streetscapes and much more. You can also learn more about our work being presented in the list below (Google affiliations highlighted in blue).
Tutorials Designing Learning Dynamics Organizers: Marta Garnelo, David Balduzzi, Wojciech Czarnecki Where Neuroscience meets AI (And What’s in Store for the Future) Organizers: Jane Wang, Kevin Miller, Adam Marblestone Offline Reinforcement Learning: From Algorithm Design to Practical Applications Organizers: Sergey Levine, Aviral Kumar Practical Uncertainty Estimation and Out-of-Distribution Robustness in Deep Learning Organizers: Dustin Tran, Balaji Lakshminarayanan, Jasper Snoek Abstraction & Reasoning in AI systems: Modern Perspectives Organizers: Francois Chollet, Melanie Mitchell, Christian Szegedy Policy Optimization in Reinforcement Learning Organizers: Sham M Kakade, Martha White, Nicolas Le Roux Federated Learning and Analytics: Industry Meets Academia Organizers: Brendan McMahan, Virginia Smith, Peter Kairouz Deep Implicit Layers: Neural ODEs, Equilibrium Models, and Differentiable Optimization Organizers: David Duvenaud, J. Zico Kolter, Matthew Johnson Beyond Accuracy: Grounding Evaluation Metrics for Human-Machine Learning Systems Organizers: Praveen Chandar, Fernando Diaz, Brian St. ThomasWorkshops Black in AI Workshop @ NeurIPS 2020 (Diamond Sponsor)Mentorship Roundtables: Natasha Jacques
LatinX in AI Workshop @ NeurIPS 2020 (Platinum Sponsor)Organizers include: Pablo Samuel CastroInvited Speaker: Fernanda ViégasMentorship Roundtables: Tomas Izo
Queer in AI Workshop @ NeurIPS 2020 (Platinum Sponsor) Organizers include: Raphael Gontijo LopesWomen in Machine Learning (Platinum Sponsor)Organizers include: Xinyi Chen, Jessica SchrouffInvited Speaker: Fernanda ViégasSponsor Talk: Jessica SchrouffMentorship Roundtables: Hanie Sedghi, Marc Bellemare, Katherine Heller, Rianne van den Berg, Natalie Schluter, Colin Raffel, Azalia Mirhoseini, Emily Denton, Jesse Engel, Anusha Ramesh, Matt Johnson, Jeff Dean, Laurent Dinh, Samy Bengio, Yasaman Bahri, Corinna Cortes, Nicolas le Roux, Hugo Larochelle, Sergio Guadarrama, Natasha Jaques, Pablo Samuel Castro, Elaine Le, Cory Silvear
Muslims in ML Organizers include: Mohammad Norouzi Resistance AI Workshop Organizers include: Elliot Creager, Raphael Gontijo Lopes Privacy Preserving Machine Learning — PriML and PPML Joint Edition Organizers include: Adria Gascon, Mariana Raykova OPT2020: Optimization for Machine Learning Organizers include: Courtney Paquette Machine Learning for Health (ML4H): Advancing Healthcare for All Organizers include: Subhrajit Roy Human in the Loop Dialogue SystemsOrganizers include: Rahul GoelInvited Speaker: Ankur Parikh
Self-Supervised Learning for Speech and Audio Processing Organizers include: Tara SainathInvited Speaker: Bhuvana Ramabhadran
3rd Robot Learning Workshop Organizers include: Alex Bewley, Vincent VanhouckeInvited Speaker: Pete Florence Workshop on Deep Learning and Inverse Problems Invited Speaker: Peyman Milanfar Crowd Science Workshop: Remoteness, Fairness, and Mechanisms as Challenges of Data Supply by Humans for Automation Invited Speakers: Lora Aroyo, Praveen Paritosh Workshop on Fair AI in Finance Invited Speakers: Berk Ustun, Madeleine Clare Elish Object Representations for Learning and Reasoning Panel Moderator: Klaus Greff Deep Reinforcement Learning Organizers include: Chelsea FinnInvited Speaker: Marc Bellemare Algorithmic Fairness Through the Lens of Causality and Interpretability Organizers include: Awa Dieng, Jessica Schrouff, Fernando Diaz Machine Learning for the Developing World (ML4D) Steering Committee Member: Ernest Mwebaze Machine Learning for Engineering Modeling, Simulation and Design Organizers include: Stephan Hoyer Machine Learning for Creativity and Design Organizers include: Adam Roberts, Daphne Ippolito Invited Speaker: Jesse Engel Cooperative AI Invited Speaker: Natasha Jaques International Workshop on Scalability, Privacy, and Security in Federated Learning (SpicyFL 2020) Invited Speaker: Brendan McMahan Machine Learning for Molecules Organizers include: Jennifer WeiInvited Speaker: Benjamin Sanchez-Lengeling Navigating the Broader Impacts of AI Research Panelists include: Nyalleng Moorosi, Colin Raffel, Natalie Schluter, Ben Zevenbergen Beyond BackPropagation: Novel Ideas for Training Neural Architectures Organizers include: Yanping Huang Differentiable Computer Vision, Graphics, and Physics in Machine Learning Invited Speaker: Andrea Tagliasacchi AI for Earth Sciences Invited Speaker: Milind Tambe Machine Learning for Mobile Health Organizers include: Katherine Heller, Marianne Njifon Shared Visual Representations in Human and Machine Intelligence (SVRHM) Invited Speaker: Gamaleldin Elsayed The Challenges of Real World Reinforcement Learning Organizers include: Gabriel Dulac-ArnoldInvited Speaker: Chelsea Finn Workshop on Computer Assisted Programming (CAP) Organizers include: Charles Sutton, Augustus Odena Self-Supervised Learning — Theory and Practice Organizers include: Barret ZophInvited Speaker: Quoc V. Le Offline Reinforcement Learning Organizers include: Rishabh Agarwal, George Tucker Machine Learning for Systems Organizers include: Anna Goldie, Azalia Mirhoseini, Martin Maas Invited Speaker: Ed Chi Deep Learning Through Information Geometry Organizers include: Alexander Alemi
Expo Drifting Efficiently Through the Stratosphere Using Deep Reinforcement Learning Organizers include: Sal Candido Accelerating Eye Movement Research via Smartphone Gaze Organizers include: Vidhya Navalpakkam Mining and Learning with Graphs at Scale Organizers include: Bryan Perozzi, Vahab Mirrokni, Jonathan Halcrow, Jakub Lacki *Work performed while at Google
