Fast semantic segmentation CNNs for FPGAs

In this contribution small semantic segmentation CNNs are evaluated against traditional segmentation approaches and state of the art segmentation CNNs. The CNNs are optimized for the implementation on frame grabber FPGAs. A dataset of industrial burner flames and a dataset of transparent plastic granules is used to assess the segmentation performance of the models. VisualApplets by Basler AG is used to implement the models on an FPGA. The implemented models reach foreground IoU values of up to 96.7 %. The inference of a 552 x 552 pixel image takes slightly more than 1 ms. The latency between the start of an input line to the output of the line is 0.1 to 1.9 ms for streaming an 8192 pixel wide image.