Fig. 1. Flow chart of the proposed model

Fig. 2. Joint training U-Net structure

Fig. 3. Structure of the attention mechanism

Fig. 4. Semantic segmentation results

Fig. 5. Low light enhancement results from different methods in LISU datasets

Fig. 6. Comparison of the visualization results with and without semantic informations

Fig. 7. Comparison of the attention map visualization with and without semantic informations

Fig. 8. Comparison of the visualization results with and without attention mechanism module