ENet-Label-Torch is available now (a light-weight and effective lane detection model)
cardwing opened this issue · comments
Our ENet-Label-Torch has been released. More details can be found in my repo.
Key features:
(1) ENet-label is a light-weight lane detection model based on ENet and adopts self attention distillation (more details can be found in our paper which will be published soon).
(2) It has 20 × fewer parameters and runs 10 × faster compared to the state-of-the-art SCNN, and achieves 72.0 (F1-measure) on CULane testing set (better than SCNN which achieves 71.6).
(Do not hesitate to try our model!!!)
Performance on CULane testing set (F1-measure):
| Category | SCNN-Torch | SCNN-Tensorflow | ENet-Label-Torch |
|---|---|---|---|
| Normal | 90.6 | 90.2 | 90.7 |
| Crowded | 69.7 | 71.9 | 70.8 |
| Night | 66.1 | 64.6 | 65.9 |
| No line | 43.4 | 45.8 | 44.7 |
| Shadow | 66.9 | 73.8 | 70.6 |
| Arrow | 84.1 | 83.8 | 85.8 |
| Dazzle light | 58.5 | 59.5 | 64.4 |
| Curve | 64.4 | 63.4 | 65.4 |
| Crossroad | 1990 | 4137 | 2729 |
| Total | 71.6 | 71.3 | 72.0 |
| Runtime(ms) | 133.5 | -- | 13.4 |
| Parameter(M) | 20.72 | -- | 0.98 |
Hi, We do not hesitate to try your model, But lua and torch will stop many people. That will be great if there is pytorch and python version. All in all, thank you again.
@mengzhibin @cardwing Things will greater if there is tensor flow version:)
I will implement both Pytorch version and TensorFlow version when I am free. However, you can just read the torch codes, then write the PyTorch / TensorFlow codes and pull request in my Github repo. I will respond to you soon.
@mengzhibin Hi! May I recommend my repo? Both SCNN and LaneNet are state of art lane detection cnn model. I implement them here and here in pytorch. Hope this helps. Thanks!
Thanks for your work, that's will help a lot. @harryhan618
@mengzhibin ERFNet-CULane-PyTorch has been released. (It can achieve 73.1 F1-measure in CULane testing set)
@cardwing Thanks for your work of lane detection. I have checked the code.
@mengzhibin You are welcome. And you can add other features, e.g., SAD, to the PyTorch codes. You can obtain better performance in all benchmarks.