SparseMask: Differentiable Connectivity Learning for Dense Image Prediction
This work addresses the need for more efficient and adaptable neural network architectures in computer vision tasks like segmentation, though it is incremental as it builds on existing encoder-decoder frameworks.
The paper tackled the problem of automatically designing efficient network architectures for dense image prediction by learning sparse connectivity in decoders, achieving competitive segmentation results with over three times faster speed and less than half the parameters compared to state-of-the-art methods.
In this paper, we aim at automatically searching an efficient network architecture for dense image prediction. Particularly, we follow the encoder-decoder style and focus on designing a connectivity structure for the decoder. To achieve that, we design a densely connected network with learnable connections, named Fully Dense Network, which contains a large set of possible final connectivity structures. We then employ gradient descent to search the optimal connectivity from the dense connections. The search process is guided by a novel loss function, which pushes the weight of each connection to be binary and the connections to be sparse. The discovered connectivity achieves competitive results on two segmentation datasets, while runs more than three times faster and requires less than half parameters compared to the state-of-the-art methods. An extensive experiment shows that the discovered connectivity is compatible with various backbones and generalizes well to other dense image prediction tasks.