search

mit-han-lab / efficientvit

EfficientViT is a new family of vision models for efficient high-resolution vision.
Apache License 2.0
1.79k stars 164 forks source link

Segmentation training code #35

Closed d710055071 closed 12 months ago

d710055071 commented 12 months ago

I cannot reproduce the semantic segmentation results. When can I open up the training code for semantic segmentation? Thank you very much

d710055071 commented 12 months ago

+---------------+-------+-------+--------+-----------+--------+-------+ | Class | IoU | Acc | Fscore | Precision | Recall | Dice | +---------------+-------+-------+--------+-----------+--------+-------+ | road | 0.0 | 0.0 | nan | nan | 0.0 | 0.0 | | sidewalk | 23.59 | 93.33 | 38.18 | 24.0 | 93.33 | 38.18 | | building | 62.84 | 78.81 | 77.18 | 75.62 | 78.81 | 77.18 | | wall | 21.46 | 23.71 | 35.34 | 69.35 | 23.71 | 35.34 | | fence | 2.27 | 2.43 | 4.45 | 25.82 | 2.43 | 4.45 | | pole | 1.34 | 1.39 | 2.65 | 29.64 | 1.39 | 2.65 | | traffic light | 3.95 | 29.46 | 7.6 | 4.37 | 29.46 | 7.6 | | traffic sign | 88.65 | 95.64 | 93.98 | 92.38 | 95.64 | 93.98 | | vegetation | 57.61 | 79.86 | 73.1 | 67.4 | 79.86 | 73.1 | | terrain | 21.44 | 28.63 | 35.31 | 46.07 | 28.63 | 35.31 | | sky | 0.0 | 0.0 | nan | 0.0 | 0.0 | 0.0 | | person | 72.05 | 95.72 | 83.76 | 74.45 | 95.72 | 83.76 | | rider | 18.65 | 21.77 | 31.44 | 56.57 | 21.77 | 31.44 | | car | 18.36 | 26.03 | 31.02 | 38.39 | 26.03 | 31.02 | | truck | 29.98 | 30.31 | 46.13 | 96.51 | 30.31 | 46.13 | | bus | 31.49 | 53.16 | 47.9 | 43.58 | 53.16 | 47.9 | | train | nan | nan | nan | nan | nan | nan | | motorcycle | 27.58 | 36.57 | 43.24 | 52.88 | 36.57 | 43.24 | | bicycle | 0.0 | 0.0 | nan | 0.0 | 0.0 | 0.0 | +---------------+-------+-------+--------+-----------+--------+-------+ 2023-09-30 11:13:01,991 - mmseg - INFO - Summary: 2023-09-30 11:13:01,991 - mmseg - INFO - +-------+-------+-------+---------+------------+---------+-------+ | aAcc | mIoU | mAcc | mFscore | mPrecision | mRecall | mDice | +-------+-------+-------+---------+------------+---------+-------+ | 85.45 | 26.74 | 38.71 | 43.42 | 46.88 | 38.71 | 36.18 | +-------+-------+-------+---------+------------+---------+-------+ 2023-09-30 11:13:01,994 - mmseg - INFO - The previous best checkpoint /data2/dongzf/workdir/efficentvit_cityspace_v1/best_mIoU_epoch_693.pth was removed 2023-09-30 11:13:02,067 - mmseg - INFO - Now best checkpoint is saved as best_mIoU_epoch_697.pth. 2023-09-30 11:13:02,067 - mmseg - INFO - Best mIoU is 0.2674 at 697 epoch.