search

shirokunet / lane_segmentation

Robust Highway Lane Segmentation Based on LaneNet Trained BDD100K
29 stars 13 forks source link

Pre-trained model weights : 2022 #3

Open MihirD97 opened 1 year ago

MihirD97 commented 1 year ago

Hello, I was trying to implement the script 02-LaneNet-detection on the bdd100k dataset. However, when I try to restore the weights, specifically in the LaneNet class - `

restore

    saver = tf.compat.v1.train.Saver()
    saver.restore(sess=self.sess, save_path=weights_path)`

It throws me the following error -

NotFoundError: 2 root error(s) found. (0) NOT_FOUND: Key lanenet_model/inference/decode/deconv_1/deconv_1/kernel not found in checkpoint [[{{node save/RestoreV2}}]] [[save/RestoreV2/_79]] (1) NOT_FOUND: Key lanenet_model/inference/decode/deconv_1/deconv_1/kernel not found in checkpoint [[{{node save/RestoreV2}}]] 0 successful operations. 0 derived errors ignored.

These were some of the other messages that were raised in the same error-

RuntimeError: Key _CHECKPOINTABLE_OBJECT_GRAPH not found in checkpoint

During handling of the above exception, another exception occurred:

I downloaded the weights from the LaneNet github repo - https://github.com/MaybeShewill-CV/lanenet-lane-detection This is the direct link - https://www.dropbox.com/sh/0b6r0ljqi76kyg9/AADedYWO3bnx4PhK1BmbJkJKa?dl=0

I think this might be because the weights uploaded on the LaneNet github are updated wrt this implementation.

Could anyone please share the weights file used in this implementation? Specifically the one mentioned in this line -

weights_path = 'model/tusimple_lanenet/tusimple_lanenet_vgg_2018-10-19-13-33-56.ckpt-200000'

Thank you