Classification

Implement models for multi-classes, multi-labels classification tasks

1. Architectures

• MobileNet V1: Efficient Convolutional Neural Networks for Mobile Vision Applications
• MobileNet V2: Inverted Residuals and Linear Bottlenecks
• MobileNet V3: Searching for MobileNetV3
• EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks
• PP-LCNet: : A Lightweight CPU Convolutional Neural Network
• CoAtNet: Marrying Convolution and Attention for All Data Sizes
• An image is worth 16X16 words: transformers for image recognition at scale
• Swin Transformer: Hierarchical Vision Transformer using Shifted Windows

2. Criteria and Metric

• Cross entropy
• Focal Loss
• PolyLoss: A Polynomial Expansion Perspective of Loss Functions
• Accuracy, Recall, Precision, F1-Score
• FAR (False Acception Rate)
• FRR (False Rejection Rate)

3. Usage

Training

CUDA_VISIBLE_DEVICES=0,1,2,3 python main.py configs/MNIST/training.yaml --num-epochs 20 --gpu-indices 0,1,2,3

Testing

CUDA_VISIBLE_DEVICES=0,1,2,3 python main.py configs/MNIST/testing.yaml --gpu-indices 0,1,2,3 --checkpoint-path <str>

Resume

CUDA_VISIBLE_DEVICES=0,1,2,3 python main.py --config config/MNIST/training.yaml --num-epoch 20 --num-gpus 0,1,2,3 --resume-path <str>

Note. Run TensorBoard on Server

You have to create a ssh connection using port forwarding:

ssh -L 16006:127.0.0.1:6006 user@host

Then you run the tensorboard command:

tensorboard --logdir=/path/to/logs

Then you can easily access the tensorboard in your browser under:

localhost:16006/

4. Techniques

1. Enhance performace by applying mixup

   References:

   • [1]mixup: Beyond Empirical Risk Minimization
   • [2] mixup-cifar10
   • [3] Enhancing Neural Networks with Mixup in Pytorch

2. Pruning Techniques

3. Semi-supervised Learning

4. Unsupervised Learning

5. Self-supervised Learning