Image classification using CNN
This is a part of the dataset in Competition 2 - Shopee Code League. The trainning include 10 image categories, around 1500 trainning images in each category clean from the original dataset. The 10 category names as follows:
00: Maxi dress
01: Muslim dress
02: T-shirt
03: Hoodie shirt
04: Jean
05: Ring
06: Ear ring
07: Cap
08: Wallet
09: Bag
10: Phone case
All of the images are real images on Shopee platform - the leading e-commerce online shopping platform in Southeast Asia and Taiwan.
https://drive.google.com/file/d/10xAzLNwr1ye1KeCyAskmrzvVELOlwQ6U/view?usp=sharing
The feature of each image is extracted by using pre-trained model ResNet50, with weights are imagenet.
After feature extracted by ResNet50, the model will go through a 1500 nodes hidden layer before go to the output layer. The nodes number can modify to achieve better accuracy.
Run c2_train_feature_extraction.py
Run c2_test_feature_extraction.py
Run c2_training.py
Run c2_predict.py
The trainning accuracy is reached at 0.9437 after 5 epochs.
The testing accuracy is reached at 0.81
