This project is a homework of Parallel Computing and Deep Learning System course. Using VGG16 to clasify 8 defects on Chip Panel. The main purpose of the project is to measure the time execution of each layer by using TimeHooked (a function from Chainer platform).
Requirements:
- Chainer
- Cupy
To run the training code: cd ./path~ python hw2_train.py --dataset [linktodataset] --out [foldercontainmodel] --unit [outputclass] --batchsize [sizeofbatch] --epoch [numberofepoch]
To run the inference code:
- Download Test Data: Link: https://drive.google.com/drive/folders/18k1W-dUwal-zJJ0EEmfs5wl0URTEQmeg?usp=sharing
- Test data folder: ~/data/test/
- Trained model: ~/weights/ADC.model
- GPU: cd ./path~ python hw2_inference.py --dataset [linktodataset] --out [foldercontainmode] --unit [outputclass] --gpu True
- CPU: cd ./path~ python hw2_inference.py --dataset [linktodataset] --out [foldercontainmode] --unit [outputclass]
Results:
