SuperResolution

SuperResolution is a model that transforms low-resolution images into high-resolution images through deep learning or other algorithms, aimed at enhancing image details and visual quality.

1. Manually download the model and upload it to raspberrypi5, or pull the model repository via the following command.

git clone https://huggingface.co/AXERA-TECH/SuperResolution

File Description:

m5stack@raspberrypi:~/rsp/SuperResolution $ ls -lh
total 20K
-rw-rw-r-- 1 m5stack m5stack 3.8K Sep  4 18:49 config.json
drwxrwxr-x 4 m5stack m5stack 4.0K Sep  4 18:49 model_convert
drwxrwxr-x 5 m5stack m5stack 4.0K Sep  4 19:12 python
-rw-rw-r-- 1 m5stack m5stack 2.3K Sep  4 18:49 README.md
drwxrwxr-x 2 m5stack m5stack 4.0K Sep  4 19:03 video

2. Create a virtual environment

   python -m venv sr

   python -m venv sr

3. Activate the virtual environment

   source sr/bin/activate

   source sr/bin/activate

4. Install dependencies

   pip install https://github.com/AXERA-TECH/pyaxengine/releases/download/0.1.3.rc1/axengine-0.1.3-py3-none-any.whl
   pip install opencv-python torch torchvision tqdm scikit-image

   pip install https://github.com/AXERA-TECH/pyaxengine/releases/download/0.1.3.rc1/axengine-0.1.3-py3-none-any.whl pip install opencv-python torch torchvision tqdm scikit-image

5. Run

   python python/run_axmodel.py --model model_convert/axmodel/edsr_baseline_x2_1.axmodel --dir_demo video/test_1920x1080.mp4

   python python/run_axmodel.py --model model_convert/axmodel/edsr_baseline_x2_1.axmodel --dir_demo video/test_1920x1080.mp4

Execution result:

(sr) m5stack@raspberrypi:~/rsp/SuperResolution $ python python/run_axmodel.py --model model_convert/axmodel/edsr_baseline_x2_1.axmodel --dir_demo video/test_1920x1080.mp4 
[INFO] Available providers:  ['AXCLRTExecutionProvider']
[INFO] Using provider: AXCLRTExecutionProvider
[INFO] SOC Name: AX650N
[INFO] VNPU type: VNPUType.DISABLED
[INFO] Compiler version: 4.2 6bff2f67
100%|█████████████████████████████████████████| 267/267 [06:21<00:00,  1.43s/it]
Total time: 275.618 seconds for 267 frames
Average time: 1.032 seconds for each frame

Original video screenshot:

Output video screenshot: