metadata

title: SpecDETR Gradio App
emoji: 🚀
colorFrom: indigo
colorTo: blue
sdk: gradio
sdk_version: 3.50.2
app_file: app.py
pinned: false

Check out the configuration reference at https://huggingface.co./docs/hub/spaces-config-reference

SpecDETR Hugging Face Space

This repository contains a Gradio demo for SpecDETR, supporting both image and video detection.

Usage

Upload your model config (.py) and checkpoint (.pth) files to the interface.
Test detection on images or videos using the Gradio UI.

Files

app.py: Gradio app entry point.
requirements.txt: Python dependencies for Hugging Face Spaces.
download_spod.py: (Optional) Script for dataset download and config update.

Deployment

Push this repo to your Hugging Face Space.
The Space will auto-build and launch the Gradio demo.

<<<<<<< HEAD ## Note - If you need to add default configs or checkpoints, place them in the repo and update `app.py` accordingly.

The code is built upon mmdetection 3.0.0 and runs on the Ubuntu system.

Create a new conda environment and activate the environment. Requires Python>=3.7.
Install Pytorch. Requires torch>=1.8.

Install mmengine and mmcv.

pip install mmengine==0.7.3
pip install mmcv>=2.0.0

Clone this repository:

SpecDETR_ROOT=/path/to/clone/SpecDETR
git clone https://github.com/ZhaoxuLi123/SpecDETR $SpecDETR_ROOT

Compile and install mmdet. If mmdet>=3.0.0 is already installed in the environment, skip this step.
```
cd SpecDETR_ROOT
pip install -v -e .
```

SPOD Dataset Preparation

Download SPOD_30b_8c.zip from Baidu Netdisk (code: 1234) or Google Drive.

Unzip SPOD_30b_8c.zip to the specified path.

SPOD_ROOT=/path/to/unzip/SPOD_dataset
unzip -d SPOD_ROOT SPOD_30b_8c.zip

Update the dataset configuration file ./configs/base/datasets/hsi_detection.py
```
Line3 data_root = 'SPOD_ROOT/SPOD_30b_8c/'
```

Benchmark Evaluation and Training

Pre-trained Weights Preparation

Download the pre-trained weights file SpecDETR_100e.pth from Baidu Netdisk (code: 1234) or Google Drive.
Place the file SpecDETR_100e.pth under ./work_dirs/SpecDETR/

Evaluation

Obtain detection results and evaluate AP.
```
python test.py
```
Evaluate inference speed FPS.
```
python train.py
```
Evaluate flops.
```
python benchmark.py
```
Retrain the network from scratch.
```
python train.py
```

Note: Although we have fixed all random seeds, there may still be slight differences in AP performance after retraining from scratch each time. This difference originates from the underlying mechanism of CUDA.

Citation

If the work or the code is helpful, please cite the paper:

@article{li2024specdetr,
  title={SpecDETR: A Transformer-based Hyperspectral Point Object Detection Network},
  author={Li, Zhaoxu and An, Wei and Guo, Gaowei and Wang, Longguang and Wang, Yingqian and Lin, Zaiping},
  journal={arXiv preprint arXiv:2405.10148},
  year={2024}
}

Contact

For further questions or details, please directly reach out to [email protected].

dc3c711 (Update README.md)