In this work, we propose a novel scale-aware progressive training mechanism to address large scale variations across faces.
We observe that these proposed schemes are capable of facilitating the learning of discriminative feature representations.
To this end, we comprehensively investigate three types of ranking constraints, i. e., global ranking, class-specific ranking and IoU-guided ranking losses.
Particularly, with the same architecture of PSPNet (ResNet-18), our method outperforms the single-dataset baseline by 5. 65\%, 6. 57\%, and 5. 79\% of mIoU on the validation sets of Cityscapes, BDD100K, CamVid, respectively.
Such bin regularization (BR) mechanism encourages the weight distribution of each quantization bin to be sharp and approximate to a Dirac delta distribution ideally.
On the Xilinx ZU2 @330 MHz and ZU9 @330 MHz, we achieve equivalently state-of-the-art performance on our benchmarks by na\"ive implementations without optimizations, and the throughput is further improved up to 1. 26x by leveraging heterogeneous optimizations in DNNVM.