Data-Free Backbone Fine-Tuning for Pruned Neural Networks
This addresses a privacy and data availability issue for deploying compressed models in scenarios where training data is inaccessible, though it is incremental as it builds on existing pruning and data-free techniques.
The paper tackles the problem of fine-tuning pruned neural networks without access to original training data by using synthetically generated images and intermediate supervision to mimic the unpruned backbone's feature maps, achieving promising performance on tasks like 2D human pose estimation, object detection, and image classification.
Model compression techniques reduce the computational load and memory consumption of deep neural networks. After the compression operation, e.g. parameter pruning, the model is normally fine-tuned on the original training dataset to recover from the performance drop caused by compression. However, the training data is not always available due to privacy issues or other factors. In this work, we present a data-free fine-tuning approach for pruning the backbone of deep neural networks. In particular, the pruned network backbone is trained with synthetically generated images, and our proposed intermediate supervision to mimic the unpruned backbone's output feature map. Afterwards, the pruned backbone can be combined with the original network head to make predictions. We generate synthetic images by back-propagating gradients to noise images while relying on L1-pruning for the backbone pruning. In our experiments, we show that our approach is task-independent due to pruning only the backbone. By evaluating our approach on 2D human pose estimation, object detection, and image classification, we demonstrate promising performance compared to the unpruned model. Our code is available at https://github.com/holzbock/dfbf.