CNN Filter DB: An Empirical Investigation of Trained Convolutional Filters
This work addresses practical issues in computer vision for researchers and practitioners by providing empirical insights into CNN filter distributions, though it is incremental as it builds on existing analysis of distribution shifts.
The authors tackled the problem of understanding transferability and robustness in CNNs by analyzing shifts in learned convolutional filters, using a dataset of over 1.4 billion filters from hundreds of trained models. They found that model pre-training can succeed on arbitrary datasets meeting size and variance conditions, and identified degenerated filters that reduce robustness and fine-tuning suitability.
Currently, many theoretical as well as practically relevant questions towards the transferability and robustness of Convolutional Neural Networks (CNNs) remain unsolved. While ongoing research efforts are engaging these problems from various angles, in most computer vision related cases these approaches can be generalized to investigations of the effects of distribution shifts in image data. In this context, we propose to study the shifts in the learned weights of trained CNN models. Here we focus on the properties of the distributions of dominantly used 3x3 convolution filter kernels. We collected and publicly provide a dataset with over 1.4 billion filters from hundreds of trained CNNs, using a wide range of datasets, architectures, and vision tasks. In a first use case of the proposed dataset, we can show highly relevant properties of many publicly available pre-trained models for practical applications: I) We analyze distribution shifts (or the lack thereof) between trained filters along different axes of meta-parameters, like visual category of the dataset, task, architecture, or layer depth. Based on these results, we conclude that model pre-training can succeed on arbitrary datasets if they meet size and variance conditions. II) We show that many pre-trained models contain degenerated filters which make them less robust and less suitable for fine-tuning on target applications. Data & Project website: https://github.com/paulgavrikov/cnn-filter-db