https://github.com/vievie31/isc21-onnx

Exporting as ONNX model the 1st Place Solution of the Facebook AI Image Similarity Challenge (ISC21) : Descriptor Track.
https://github.com/vievie31/isc21-onnx

Last synced: about 4 hours ago
JSON representation

Exporting as ONNX model the 1st Place Solution of the Facebook AI Image Similarity Challenge (ISC21) : Descriptor Track.

• Host: GitHub
• URL: https://github.com/vievie31/isc21-onnx
• Owner: VieVie31
• Created: 2024-06-29T16:06:43.000Z (5 months ago)
• Default Branch: main
• Last Pushed: 2024-06-30T12:20:49.000Z (5 months ago)
• Last Synced: 2024-07-19T16:53:47.029Z (4 months ago)
• Language: Python
• Size: 88.9 KB
• Stars: 0
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# ISC21-ONNX

Exporting as ONNX model  the 1st Place Solution of the Facebook AI Image Similarity Challenge (ISC21) : Descriptor Track.

## Exporting to ONNX

Go in `isc21_onnx` folder. 

Then run: `export_to_onnx.py` to export the model into onnx.

Finaly use: `compare_models.py` to check that the onnx model produce outputs similar to original pytorch one.

The exported model expect inputs to be `512x512` images.

## Static Quantization

1. Download the dataset for the calibration ([this one for example](https://www.kaggle.com/datasets/rhtsingh/130k-images-512x512-universal-image-embeddings?resource=download))

2. Run the pre-processing for quantization `python -m onnxruntime.quantization.preprocess --input isc_ft_v107.onnx --output isc_ft_v107_prep.onnx`

3. Run the quantization script `python quantize.py --input_model isc_ft_v107_prep.onnx --output_model isc_ft_v107_quant_120.onnx --nb_images=120`

Output of the script should look something like:

```bash

(120, 1, 3, 512, 512)

DataReader created, starting calibration...

Collecting tensor data and making histogram ...

Finding optimal threshold for each tensor using 'percentile' algorithm ...

Number of tensors : 484

Number of histogram bins : 2048

Percentile : (0.0010000000000047748,99.999)

Calibrated and quantized model saved.

benchmarking fp32 model...

260.36ms

262.41ms

261.08ms

261.11ms

265.63ms

267.71ms

259.35ms

262.94ms

266.17ms

270.61ms

Avg: 263.74ms

benchmarking int8 model...

149.71ms

149.84ms

149.03ms

148.88ms

151.16ms

148.35ms

148.37ms

147.65ms

149.61ms

149.35ms

Avg: 149.20ms

```