base on A distilled Segment Anything (SAM) model capable of running real-time with NVIDIA TensorRT <h1 align="center"><span>NanoSAM</span></h1> <p align="center"><a href="#usage"/>👍 Usage</a> - <a href="#performance"/>⏱️ Performance</a> - <a href="#setup">🛠️ Setup</a> - <a href="#examples">🤸 Examples</a> - <a href="#training">🏋️ Training</a> <br>- <a href="#evaluation">🧐 Evaluation</a> - <a href="#acknowledgement">👏 Acknowledgment</a> - <a href="#see-also">🔗 See also</a></p> NanoSAM is a [Segment Anything (SAM)](https://github.com/facebookresearch/segment-anything) model variant that is capable of running in 🔥 ***real-time*** 🔥 on [NVIDIA Jetson Orin Platforms](https://store.nvidia.com/en-us/jetson/store) with [NVIDIA TensorRT](https://developer.nvidia.com/tensorrt). <!-- <img src="assets/tshirt_gif_compressed_v2.gif" height="20%" width="20%"/> --> <p align="center"><img src="assets/basic_usage_out.jpg" height="256px"/></p> <!--<img src="assets/mouse_gif_compressed.gif" height="50%" width="50%"/> --> > NanoSAM is trained by distilling the [MobileSAM](https://github.com/ChaoningZhang/MobileSAM) image encoder > on unlabeled images. For an introduction to knowledge distillation, we recommend checking out [this tutorial](https://github.com/NVIDIA-AI-IOT/jetson-intro-to-distillation). <a id="usage"></a> ## 👍 Usage Using NanoSAM from Python looks like this ```python3 from nanosam.utils.predictor import Predictor predictor = Predictor( image_encoder="data/resnet18_image_encoder.engine", mask_decoder="data/mobile_sam_mask_decoder.engine" ) image = PIL.Image.open("dog.jpg") predictor.set_image(image) mask, _, _ = predictor.predict(np.array([[x, y]]), np.array([1])) ``` <details> <summary>Notes</summary> The point labels may be | Point Label | Description | |:--------------------:|-------------| | 0 | Background point | | 1 | Foreground point | | 2 | Bounding box top-left | | 3 | Bounding box bottom-right | </details> > Follow the instructions below for how to build the engine files. <a id="performance"></a> ## ⏱️ Performance NanoSAM runs real-time on Jetson Orin Nano. <table style="border-top: solid 1px; border-left: solid 1px; border-right: solid 1px; border-bottom: solid 1px"> <thead> <tr> <th rowspan=2 style="text-align: center; border-right: solid 1px">Model †</th> <th colspan=2 style="text-align: center; border-right: solid 1px">:stopwatch: Jetson Orin Nano (ms)</th> <th colspan=2 style="text-align: center; border-right: solid 1px">:stopwatch: Jetson AGX Orin (ms)</th> <th colspan=4 style="text-align: center; border-right: solid 1px"> :dart: Accuracy (mIoU) ‡</th> </tr> <tr> <th style="text-align: center; border-right: solid 1px">Image Encoder</th> <th style="text-align: center; border-right: solid 1px">Full Pipeline</th> <th style="text-align: center; border-right: solid 1px">Image Encoder</th> <th style="text-align: center; border-right: solid 1px">Full Pipeline</th> <th style="text-align: center; border-right: solid 1px">All</th> <th style="text-align: center; border-right: solid 1px">Small</th> <th style="text-align: center; border-right: solid 1px">Medium</th> <th style="text-align: center; border-right: solid 1px">Large</th> </tr> </thead> <tbody> <tr> <td style="text-align: center; border-right: solid 1px">MobileSAM</td> <td style="text-align: center; border-right: solid 1px">TBD</td> <td style="text-align: center; border-right: solid 1px">146</td> <td style="text-align: center; border-right: solid 1px">35</td> <td style="text-align: center; border-right: solid 1px">39</td> <td style="text-align: center; border-right: solid 1px">0.728</td> <td style="text-align: center; border-right: solid 1px">0.658</td> <td style="text-align: center; border-right: solid 1px">0.759</td> <td style="text-align: center; border-right: solid 1px">0.804</td> </tr> <tr> <td style="text-align: center; border-right: solid 1px">NanoSAM (ResNet18)</td> <td style="text-align: center; border-right: solid 1px">TBD</td> <td style="text-align: center; border-right: solid 1px">27</td> <td style="text-align: center; border-right: solid 1px">4.2</td> <td style="text-align: center; border-right: solid 1px">8.1</td> <td style="text-align: center; border-right: solid 1px">0.706</td> <td style="text-align: center; border-right: solid 1px">0.624</td> <td style="text-align: center; border-right: solid 1px">0.738</td> <td style="text-align: center; border-right: solid 1px">0.796</td> </tr> </tbody> </table> <details> <summary>Notes</summary> † The MobileSAM image encoder is optimized with FP32 precision because it produced erroneous results when built for FP16 precision with TensorRT. The NanoSAM image encoder is built with FP16 precision as we did not notice a significant accuracy degredation. Both pipelines use the same mask decoder which is built with FP32 precision. For all models, the accuracy reported uses the same model configuration used to measure latency. ‡ Accuracy is computed by prompting SAM with ground-truth object bounding box annotations from the COCO 2017 validation dataset. The IoU is then computed between the mask output of the SAM model for the object and the ground-truth COCO segmentation mask for the object. The mIoU is the average IoU over all objects in the COCO 2017 validation set matching the target object size (small, medium, large). </details> <a id="setup"></a> ## 🛠️ Setup NanoSAM is fairly easy to get started with. 1. Install the dependencies 1. Install PyTorch 2. Install [torch2trt](https://github.com/NVIDIA-AI-IOT/torch2trt) 3. Install NVIDIA TensorRT 4. (optional) Install [TRTPose](https://github.com/NVIDIA-AI-IOT/trt_pose) - For the pose example. ```bash git clone https://github.com/NVIDIA-AI-IOT/trt_pose cd trt_pose python3 setup.py develop --user ``` 5. (optional) Install the Transformers library - For the OWL ViT example. ```bash python3 -m pip install transformers ``` 2. Install the NanoSAM Python package ```bash git clone https://github.com/NVIDIA-AI-IOT/nanosam cd nanosam python3 setup.py develop --user ``` 3. Build the TensorRT engine for the mask decoder 1. Export the MobileSAM mask decoder ONNX file (or download directly from [here](https://drive.google.com/file/d/1jYNvnseTL49SNRx9PDcbkZ9DwsY8up7n/view?usp=drive_link)) ```bash python3 -m nanosam.tools.export_sam_mask_decoder_onnx \ --model-type=vit_t \ --checkpoint=assets/mobile_sam.pt \ --output=data/mobile_sam_mask_decoder.onnx ``` 2. Build the TensorRT engine ```bash trtexec \ --onnx=data/mobile_sam_mask_decoder.onnx \ --saveEngine=data/mobile_sam_mask_decoder.engine \ --minShapes=point_coords:1x1x2,point_labels:1x1 \ --optShapes=point_coords:1x1x2,point_labels:1x1 \ --maxShapes=point_coords:1x10x2,point_labels:1x10 ``` > This assumes the mask decoder ONNX file is downloaded to ``data/mobile_sam_mask_decoder.onnx`` <details> <summary>Notes</summary> This command builds the engine to support up to 10 keypoints. You can increase this limit as needed by specifying a different max shape. </details> 4. Build the TensorRT engine for the NanoSAM image encoder 1. Download the image encoder: [resnet18_image_encoder.onnx](https://drive.google.com/file/d/14-SsvoaTl-esC3JOzomHDnI9OGgdO2OR/view?usp=drive_link) 2. Build the TensorRT engine ```bash trtexec \ --onnx=data/resnet18_image_encoder.onnx \ --saveEngine=data/resnet18_image_encoder.engine \ --fp16 ``` 5. Run the basic usage example ``` python3 examples/basic_usage.py \ --image_encoder=data/resnet18_image_encoder.engine \ --mask_decoder=data/mobile_sam_mask_decoder.engine ``` > This outputs a result to ``data/basic_usage_out.jpg`` That's it! From there, you can read the example code for examples on how to use NanoSAM with Python. Or try running the more advanced examples below. <a id="examples"></a> ## 🤸 Examples NanoSAM can be applied in many creative ways. ### Example 1 - Segment with bounding box <img src="assets/basic_usage_out.jpg" height="256"/> This example uses a known image with a fixed bounding box to control NanoSAM segmentation. To run the example, call ```python3 python3 examples/basic_usage.py \ --image_encoder="data/resnet18_image_encoder.engine" \ --mask_decoder="data/mobile_sam_mask_decoder.engine" ``` ### Example 2 - Segment with bounding box (using OWL-ViT detections) <img src="assets/owl_out.png" height="256"/> This example demonstrates using OWL-ViT to detect objects using a text prompt(s), and then segmenting these objects using NanoSAM. To run the example, call ```bash python3 examples/segment_from_owl.py \ --prompt="A tree" \ --image_encoder="data/resnet18_image_encoder.engine" \ --mask_decoder="data/mobile_sam_mask_decoder.engine ``` <details> <summary>Notes</summary> - While OWL-ViT does not run real-time on Jetson Orin Nano (3sec/img), it is nice for experimentation as it allows you to detect a wide variety of objects. You could substitute any other real-time pre-trained object detector to take full advantage of NanoSAM's speed. </details> ### Example 3 - Segment with keypoints (offline using TRTPose detections) <img src="assets/pose_out.png" height="256"/> This example demonstrates how to use human pose keypoints from [TRTPose](https://github.com/NVIDIA-AI-IOT/trt_pose) to control NanoSAM segmentation. To run the example, call ```bash python3 examples/segment_from_pose.py ``` This will save an output figure to ``data/segment_from_pose_out.png``. ### Example 4 - Segment with keypoints (online using TRTPose detections) <img src="assets/tshirt_gif_compressed_v2.gif" height="40%" width="40%"/> This example demonstrates how to use human pose to control segmentation on a live camera feed. This example requires an attached display and camera. To run the example, call ```python3 python3 examples/demo_pose_tshirt.py ``` ### Example 5 - Segment and track (experimental) <img src="assets/mouse_gif_compressed.gif" height="40%" width="40%"/> This example demonstrates a rudimentary segmentation tracking with NanoSAM. This example requires an attached display and camera. To run the example, call ```python3 python3 examples/demo_click_segment_track.py <image_encoder_engine> <mask_decoder_engine> ``` Once the example is running **double click** an object you want to track. <details> <summary>Notes</summary> This tracking method is very simple and can get lost easily. It is intended to demonstrate creative ways you can use NanoSAM, but would likely be improved with more work. </details> <a id="training"></a> ## 🏋️ Training You can train NanoSAM on a single GPU 1. Download and extract the COCO 2017 train images ```bash # mkdir -p data/coco # uncomment if it doesn't exist mkdir -p data/coco cd data/coco wget http://images.cocodataset.org/zips/train2017.zip unzip train2017.zip cd ../.. ``` 2. Build the MobileSAM image encoder (used as teacher model) 1. Export to ONNX ```bash python3 -m nanosam.tools.export_sam_image_encoder_onnx \ --checkpoint="assets/mobile_sam.pt" \ --output="data/mobile_sam_image_encoder_bs16.onnx" \ --model_type=vit_t \ --batch_size=16 ``` 2. Build the TensorRT engine with batch size 16 ```bash trtexec \ --onnx=data/mobile_sam_image_encoder_bs16.onnx \ --shapes=image:16x3x1024x1024 \ --saveEngine=data/mobile_sam_image_encoder_bs16.engine ``` 3. Train the NanoSAM image encoder by distilling MobileSAM ```bash python3 -m nanosam.tools.train \ --images=data/coco/train2017 \ --output_dir=data/models/resnet18 \ --model_name=resnet18 \ --teacher_image_encoder_engine=data/mobile_sam_image_encoder_bs16.engine \ --batch_size=16 ``` <details> <summary>Notes</summary> Once training, visualizations of progress and checkpoints will be saved to the specified output directory. You can stop training and resume from the last saved checkpoint if needed. For a list of arguments, you can type ```bash python3 -m nanosam.tools.train --help ``` </details> 4. Export the trained NanoSAM image encoder to ONNX ```bash python3 -m nanosam.tools.export_image_encoder_onnx \ --model_name=resnet18 \ --checkpoint="data/models/resnet18/checkpoint.pth" \ --output="data/resnet18_image_encoder.onnx" ``` You can then build the TensorRT engine as detailed in the getting started section. <a id="evaluation"></a> ## 🧐 Evaluation You can reproduce the accuracy results above by evaluating against COCO ground truth masks 1. Download and extract the COCO 2017 validation set. ```bash # mkdir -p data/coco # uncomment if it doesn't exist cd data/coco wget http://images.cocodataset.org/zips/val2017.zip wget http://images.cocodataset.org/annotations/annotations_trainval2017.zip unzip val2017.zip unzip annotations_trainval2017.zip cd ../.. ``` 2. Compute the IoU of NanoSAM mask predictions against the ground truth COCO mask annotation. ```bash python3 -m nanosam.tools.eval_coco \ --coco_root=data/coco/val2017 \ --coco_ann=data/coco/annotations/instances_val2017.json \ --image_encoder=data/resnet18_image_encoder.engine \ --mask_decoder=data/mobile_sam_mask_decoder.engine \ --output=data/resnet18_coco_results.json ``` > This uses the COCO ground-truth bounding boxes as inputs to NanoSAM 3. Compute the average IoU over a selected category or size ```bash python3 -m nanosam.tools.compute_eval_coco_metrics \ data/efficientvit_b0_coco_results.json \ --size="all" ``` <details> <summary>Notes</summary> For all options type ``python3 -m nanosam.tools.compute_eval_coco_metrics --help``. To compute the mIoU for a specific category id. ```bash python3 -m nanosam.tools.compute_eval_coco_metrics \ data/resnet18_coco_results.json \ --category_id=1 ``` </details> <a id="acknowledgement"></a> ## 👏 Acknowledgement This project is enabled by the great projects below. - [SAM](https://github.com/facebookresearch/segment-anything) - The original Segment Anything model. - [MobileSAM](https://github.com/ChaoningZhang/MobileSAM) - The distilled Tiny ViT Segment Anything model. <a id="see-also"></a> ## 🔗 See also - [Jetson Introduction to Knowledge Distillation Tutorial](https://github.com/NVIDIA-AI-IOT/jetson-intro-to-distillation) - For an introduction to knowledge distillation as a model optimization technique. - [Jetson Generative AI Playground](https://nvidia-ai-iot.github.io/jetson-generative-ai-playground/) - For instructions and tips for using a variety of LLMs and transformers on Jetson. - [Jetson Containers](https://github.com/dusty-nv/jetson-containers) - For a variety of easily deployable and modular Jetson Containers ", Assign "at most 3 tags" to the expected json: {"id":"1708","tags":[]} "only from the tags list I provide: [{"id":77,"name":"3d"},{"id":89,"name":"agent"},{"id":17,"name":"ai"},{"id":54,"name":"algorithm"},{"id":24,"name":"api"},{"id":44,"name":"authentication"},{"id":3,"name":"aws"},{"id":27,"name":"backend"},{"id":60,"name":"benchmark"},{"id":72,"name":"best-practices"},{"id":39,"name":"bitcoin"},{"id":37,"name":"blockchain"},{"id":1,"name":"blog"},{"id":45,"name":"bundler"},{"id":58,"name":"cache"},{"id":21,"name":"chat"},{"id":49,"name":"cicd"},{"id":4,"name":"cli"},{"id":64,"name":"cloud-native"},{"id":48,"name":"cms"},{"id":61,"name":"compiler"},{"id":68,"name":"containerization"},{"id":92,"name":"crm"},{"id":34,"name":"data"},{"id":47,"name":"database"},{"id":8,"name":"declarative-gui "},{"id":9,"name":"deploy-tool"},{"id":53,"name":"desktop-app"},{"id":6,"name":"dev-exp-lib"},{"id":59,"name":"dev-tool"},{"id":13,"name":"ecommerce"},{"id":26,"name":"editor"},{"id":66,"name":"emulator"},{"id":62,"name":"filesystem"},{"id":80,"name":"finance"},{"id":15,"name":"firmware"},{"id":73,"name":"for-fun"},{"id":2,"name":"framework"},{"id":11,"name":"frontend"},{"id":22,"name":"game"},{"id":81,"name":"game-engine "},{"id":23,"name":"graphql"},{"id":84,"name":"gui"},{"id":91,"name":"http"},{"id":5,"name":"http-client"},{"id":51,"name":"iac"},{"id":30,"name":"ide"},{"id":78,"name":"iot"},{"id":40,"name":"json"},{"id":83,"name":"julian"},{"id":38,"name":"k8s"},{"id":31,"name":"language"},{"id":10,"name":"learning-resource"},{"id":33,"name":"lib"},{"id":41,"name":"linter"},{"id":28,"name":"lms"},{"id":16,"name":"logging"},{"id":76,"name":"low-code"},{"id":90,"name":"message-queue"},{"id":42,"name":"mobile-app"},{"id":18,"name":"monitoring"},{"id":36,"name":"networking"},{"id":7,"name":"node-version"},{"id":55,"name":"nosql"},{"id":57,"name":"observability"},{"id":46,"name":"orm"},{"id":52,"name":"os"},{"id":14,"name":"parser"},{"id":74,"name":"react"},{"id":82,"name":"real-time"},{"id":56,"name":"robot"},{"id":65,"name":"runtime"},{"id":32,"name":"sdk"},{"id":71,"name":"search"},{"id":63,"name":"secrets"},{"id":25,"name":"security"},{"id":85,"name":"server"},{"id":86,"name":"serverless"},{"id":70,"name":"storage"},{"id":75,"name":"system-design"},{"id":79,"name":"terminal"},{"id":29,"name":"testing"},{"id":12,"name":"ui"},{"id":50,"name":"ux"},{"id":88,"name":"video"},{"id":20,"name":"web-app"},{"id":35,"name":"web-server"},{"id":43,"name":"webassembly"},{"id":69,"name":"workflow"},{"id":87,"name":"yaml"}]" returns me the "expected json"