Towards a Smaller Student: Capacity Dynamic Distillation for Efficient Image Retrieval
This work addresses efficiency and accuracy in image retrieval for applications like autonomous vehicles or surveillance, but it is incremental as it builds on existing knowledge distillation methods.
The paper tackles the problem of lightweight student models lacking representation capacity in knowledge distillation for efficient image retrieval by proposing a Capacity Dynamic Distillation framework, which achieves superior inference speed and accuracy, e.g., saving 67.13% parameters and 65.67% FLOPs with 2.11% higher mAP on the VeRi-776 dataset compared to state-of-the-art methods.
Previous Knowledge Distillation based efficient image retrieval methods employs a lightweight network as the student model for fast inference. However, the lightweight student model lacks adequate representation capacity for effective knowledge imitation during the most critical early training period, causing final performance degeneration. To tackle this issue, we propose a Capacity Dynamic Distillation framework, which constructs a student model with editable representation capacity. Specifically, the employed student model is initially a heavy model to fruitfully learn distilled knowledge in the early training epochs, and the student model is gradually compressed during the training. To dynamically adjust the model capacity, our dynamic framework inserts a learnable convolutional layer within each residual block in the student model as the channel importance indicator. The indicator is optimized simultaneously by the image retrieval loss and the compression loss, and a retrieval-guided gradient resetting mechanism is proposed to release the gradient conflict. Extensive experiments show that our method has superior inference speed and accuracy, e.g., on the VeRi-776 dataset, given the ResNet101 as a teacher, our method saves 67.13% model parameters and 65.67% FLOPs (around 24.13% and 21.94% higher than state-of-the-arts) without sacrificing accuracy (around 2.11% mAP higher than state-of-the-arts).