Implementing ResNet-18 and U-Net architectures for fine-grained image classification and semantic segmentation.
Evaluating deep-learning feature extractors and detectors in a recognition pipeline for a 100-person portrait dataset.
Extending SiamFC into a long-term tracker and evaluating confidence thresholds with precision, recall, and F-score.
Implementing LSTM and Transformer models for character-level text generation and comparing sequence lengths.