---
library_name: pytorch
license: other
tags:
- android
pipeline_tag: video-object-tracking

---

![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/track_anything/web-assets/model_demo.png)

# Track-Anything: Optimized for Qualcomm Devices

Track-Anything is a video based machine learning model to track an object in a video.

This is based on the implementation of Track-Anything found [here](https://github.com/gaomingqi/Track-Anything).
This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/main/src/qai_hub_models/models/track_anything) library to export with custom configurations. More details on model performance across various devices, can be found [here](#performance-summary).

Qualcomm AI Hub Models uses [Qualcomm AI Hub Workbench](https://workbench.aihub.qualcomm.com) to compile, profile, and evaluate this model. [Sign up](https://myaccount.qualcomm.com/signup) to run these models on a hosted Qualcomm® device.

## Getting Started
There are two ways to deploy this model on your device:

### Option 1: Download Pre-Exported Models

Below are pre-exported model assets ready for deployment.

| Runtime | Precision | Chipset | SDK Versions | Download |
|---|---|---|---|---|
| ONNX | float | Universal | QAIRT 2.42, ONNX Runtime 1.24.3 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/track_anything/releases/v0.50.2/track_anything-onnx-float.zip)
| TFLITE | float | Universal | QAIRT 2.43, TFLite 2.19.1 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/track_anything/releases/v0.50.2/track_anything-tflite-float.zip)

For more device-specific assets and performance metrics, visit **[Track-Anything on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/track_anything)**.


### Option 2: Export with Custom Configurations

Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/main/src/qai_hub_models/models/track_anything) Python library to compile and export the model with your own:
- Custom weights (e.g., fine-tuned checkpoints)
- Custom input shapes
- Target device and runtime configurations

This option is ideal if you need to customize the model beyond the default configuration provided here.

See our repository for [Track-Anything on GitHub](https://github.com/qualcomm/ai-hub-models/blob/main/src/qai_hub_models/models/track_anything) for usage instructions.

## Model Details

**Model Type:** Model_use_case.video_object_tracking

**Model Stats:**
- Model checkpoint: xmem
- Input resolution: 320x568
- Number of parameters (encode_key_with_shrinkage): 9.72M
- Model size (encode_key_with_shrinkage) (float): 37.1 MB
- Number of parameters (encode_value): 23.3M
- Model size (encode_value) (float): 88.8 MB
- Number of parameters (encode_key_without_shrinkage): 9.71M
- Model size (encode_key_without_shrinkage) (float): 37.1 MB
- Number of parameters (segment): 30.1M
- Model size (segment) (float): 115 MB

## Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
|---|---|---|---|---|---|---
| encode_key_with_shrinkage | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 2.38 ms | 0 - 176 MB | NPU
| encode_key_with_shrinkage | ONNX | float | Snapdragon® X2 Elite | 2.821 ms | 18 - 18 MB | NPU
| encode_key_with_shrinkage | ONNX | float | Snapdragon® X Elite | 5.593 ms | 19 - 19 MB | NPU
| encode_key_with_shrinkage | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 3.673 ms | 0 - 202 MB | NPU
| encode_key_with_shrinkage | ONNX | float | Qualcomm® QCS8550 (Proxy) | 5.163 ms | 0 - 22 MB | NPU
| encode_key_with_shrinkage | ONNX | float | Qualcomm® QCS9075 | 9.48 ms | 10 - 13 MB | NPU
| encode_key_with_shrinkage | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 2.95 ms | 0 - 165 MB | NPU
| encode_key_with_shrinkage | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 2.441 ms | 0 - 190 MB | NPU
| encode_key_with_shrinkage | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 4.078 ms | 3 - 216 MB | NPU
| encode_key_with_shrinkage | TFLITE | float | Qualcomm® QCS8275 (Proxy) | 31.185 ms | 3 - 181 MB | NPU
| encode_key_with_shrinkage | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 5.934 ms | 0 - 152 MB | NPU
| encode_key_with_shrinkage | TFLITE | float | Qualcomm® SA8775P | 9.551 ms | 3 - 182 MB | NPU
| encode_key_with_shrinkage | TFLITE | float | Qualcomm® QCS9075 | 11.381 ms | 2 - 29 MB | NPU
| encode_key_with_shrinkage | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 11.725 ms | 3 - 200 MB | NPU
| encode_key_with_shrinkage | TFLITE | float | Qualcomm® SA7255P | 31.185 ms | 3 - 181 MB | NPU
| encode_key_with_shrinkage | TFLITE | float | Qualcomm® SA8295P | 10.381 ms | 3 - 165 MB | NPU
| encode_key_with_shrinkage | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 3.281 ms | 0 - 180 MB | NPU
| encode_key_without_shrinkage | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 3.494 ms | 0 - 216 MB | NPU
| encode_key_without_shrinkage | ONNX | float | Snapdragon® X2 Elite | 3.944 ms | 64 - 64 MB | NPU
| encode_key_without_shrinkage | ONNX | float | Snapdragon® X Elite | 6.919 ms | 66 - 66 MB | NPU
| encode_key_without_shrinkage | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 5.031 ms | 61 - 272 MB | NPU
| encode_key_without_shrinkage | ONNX | float | Qualcomm® QCS8550 (Proxy) | 6.785 ms | 25 - 60 MB | NPU
| encode_key_without_shrinkage | ONNX | float | Qualcomm® QCS9075 | 11.513 ms | 62 - 65 MB | NPU
| encode_key_without_shrinkage | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 4.365 ms | 62 - 254 MB | NPU
| encode_key_without_shrinkage | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 3.567 ms | 0 - 227 MB | NPU
| encode_key_without_shrinkage | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 5.315 ms | 18 - 242 MB | NPU
| encode_key_without_shrinkage | TFLITE | float | Qualcomm® QCS8275 (Proxy) | 33.109 ms | 20 - 205 MB | NPU
| encode_key_without_shrinkage | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 7.54 ms | 8 - 16 MB | NPU
| encode_key_without_shrinkage | TFLITE | float | Qualcomm® SA8775P | 11.419 ms | 0 - 186 MB | NPU
| encode_key_without_shrinkage | TFLITE | float | Qualcomm® QCS9075 | 12.925 ms | 20 - 63 MB | NPU
| encode_key_without_shrinkage | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 14.096 ms | 20 - 223 MB | NPU
| encode_key_without_shrinkage | TFLITE | float | Qualcomm® SA7255P | 33.109 ms | 20 - 205 MB | NPU
| encode_key_without_shrinkage | TFLITE | float | Qualcomm® SA8295P | 12.764 ms | 20 - 186 MB | NPU
| encode_key_without_shrinkage | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 4.39 ms | 18 - 205 MB | NPU
| encode_value | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 6.611 ms | 13 - 179 MB | NPU
| encode_value | ONNX | float | Snapdragon® X2 Elite | 6.029 ms | 39 - 39 MB | NPU
| encode_value | ONNX | float | Snapdragon® X Elite | 12.469 ms | 38 - 38 MB | NPU
| encode_value | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 9.887 ms | 14 - 222 MB | NPU
| encode_value | ONNX | float | Qualcomm® QCS8550 (Proxy) | 15.782 ms | 0 - 45 MB | NPU
| encode_value | ONNX | float | Qualcomm® QCS9075 | 18.37 ms | 6 - 14 MB | NPU
| encode_value | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 8.416 ms | 7 - 176 MB | NPU
| encode_value | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 6.288 ms | 0 - 206 MB | NPU
| encode_value | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 8.258 ms | 3 - 285 MB | NPU
| encode_value | TFLITE | float | Qualcomm® QCS8275 (Proxy) | 51.415 ms | 3 - 218 MB | NPU
| encode_value | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 11.784 ms | 3 - 5 MB | NPU
| encode_value | TFLITE | float | Qualcomm® SA8775P | 16.955 ms | 3 - 223 MB | NPU
| encode_value | TFLITE | float | Qualcomm® QCS9075 | 18.88 ms | 0 - 54 MB | NPU
| encode_value | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 22.685 ms | 3 - 282 MB | NPU
| encode_value | TFLITE | float | Qualcomm® SA7255P | 51.415 ms | 3 - 218 MB | NPU
| encode_value | TFLITE | float | Qualcomm® SA8295P | 19.58 ms | 0 - 219 MB | NPU
| encode_value | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 7.166 ms | 3 - 203 MB | NPU
| segment | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 10.467 ms | 25 - 244 MB | NPU
| segment | ONNX | float | Snapdragon® X2 Elite | 11.785 ms | 38 - 38 MB | NPU
| segment | ONNX | float | Snapdragon® X Elite | 23.511 ms | 37 - 37 MB | NPU
| segment | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 17.397 ms | 25 - 268 MB | NPU
| segment | ONNX | float | Qualcomm® QCS8550 (Proxy) | 23.232 ms | 0 - 45 MB | NPU
| segment | ONNX | float | Qualcomm® QCS9075 | 39.003 ms | 21 - 45 MB | NPU
| segment | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 13.581 ms | 5 - 216 MB | NPU
| segment | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 10.8 ms | 2 - 226 MB | NPU
| segment | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 17.419 ms | 0 - 285 MB | NPU
| segment | TFLITE | float | Qualcomm® QCS8275 (Proxy) | 104.113 ms | 0 - 196 MB | NPU
| segment | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 23.293 ms | 2 - 4 MB | NPU
| segment | TFLITE | float | Qualcomm® SA8775P | 33.621 ms | 2 - 195 MB | NPU
| segment | TFLITE | float | Qualcomm® QCS9075 | 38.161 ms | 0 - 83 MB | NPU
| segment | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 53.173 ms | 2 - 280 MB | NPU
| segment | TFLITE | float | Qualcomm® SA7255P | 104.113 ms | 0 - 196 MB | NPU
| segment | TFLITE | float | Qualcomm® SA8295P | 38.918 ms | 2 - 204 MB | NPU
| segment | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 13.174 ms | 0 - 208 MB | NPU

## License
* The license for the original implementation of Track-Anything can be found
  [here](https://github.com/gaomingqi/Track-Anything/blob/master/LICENSE).

## References
* [Track Anything: Segment Anything Meets Videos](https://arxiv.org/abs/2304.11968)
* [Source Model Implementation](https://github.com/gaomingqi/Track-Anything)

## Community
* Join [our AI Hub Slack community](https://aihub.qualcomm.com/community/slack) to collaborate, post questions and learn more about on-device AI.
* For questions or feedback please [reach out to us](mailto:ai-hub-support@qti.qualcomm.com).