ESPNet: Towards Fast and Efficient Semantic Segmentation on the Embedded Devices This directory contains the pretrained models for ESPNet under three different settings: p=2, q=3 p=2, q=5 p=2, q=8 Models trained on the CamVid dataset espnet_p_2_q_8_camvid.pth Models trained on the CityScapes dataset espnet_p_2_q_3.pth espnet_p_2_q_5.pth espnet_p_2_q_8.pth