Bootstrapping Deep Neural Networks from Approximate Image Processing Pipelines
This addresses the data scarcity issue for researchers and practitioners in computer vision, enabling more efficient and accurate systems, though it is incremental as it builds on existing pipeline and neural network methods.
The paper tackles the problem of acquiring large labeled datasets for training deep neural networks to replace parts of image processing pipelines, proposing a workflow that uses the target pipeline to automatically generate training data without domain knowledge, resulting in trained networks achieving similar or better performance than the replaced components, with some cases showing reduced computational requirements.
Complex image processing and computer vision systems often consist of a processing pipeline of functional modules. We intend to replace parts or all of a target pipeline with deep neural networks to achieve benefits such as increased accuracy or reduced computational requirement. To acquire a large amount of labeled data necessary to train the deep neural network, we propose a workflow that leverages the target pipeline to create a significantly larger labeled training set automatically, without prior domain knowledge of the target pipeline. We show experimentally that despite the noise introduced by automated labeling and only using a very small initially labeled data set, the trained deep neural networks can achieve similar or even better performance than the components they replace, while in some cases also reducing computational requirements.