search

MaybeShewill-CV / lanenet-lane-detection

Unofficial implemention of lanenet model for real time lane detection
Apache License 2.0
2.34k stars 885 forks source link

Paper Sources #215

Closed LukasProgram closed 5 years ago

LukasProgram commented 5 years ago

Hi @MaybeShewill-CV

I used your network and train my own data on it. It worked pretty well. Now I want some more information about LaneNet. I found a paper from Ze Wang, Weiqiang Ren, and Qiang Qiu -> "LaneNet: Real-Time Lane Detection Networks for Autonomous Driving" is it the same LaneNet, or is it related to the LaneNet described in "Towards End-to-End Lane Detection: an Instance Segmentation Approach"?

Can you help me to understand it because I want to know how it works? They are a lot of LaneNet papers but I cannot distinguish between the LaneNet used in your Implementation from the other LaneNet described in other papers.

I want to cite your work and I thank you in advance for any answer.

MaybeShewill-CV commented 5 years ago

@LukasProgram The repo implements the paper "Towards End-to-End Lane Detection: an Instance Segmentation Approach". As for method details you may refer to the paper for details:). Btw could you please share some test result on your own dataset here if possible? :)

LukasProgram commented 5 years ago

Hi @MaybeShewill-CV
I can share the results from tensorboard if you want. I also did the validation step.

I am trying now to run the model on a smartphone, so I need to run the python3 export_inference_graph.py \ --input_type --pipeline_config_path --trained_checkpoint_prefix --output_directory

I don't know what argument should I use for pipeline_config_path. Can you help me?

MaybeShewill-CV commented 5 years ago

@LukasProgram Sorry I do not know what it is:)

LukasProgram commented 5 years ago

Ok, and what output nodes have the model or graph?

I try to run the script in https://stackoverflow.com/questions/45864363/tensorflow-how-to-convert-meta-data-and-index-model-files-into-one-graph-pb

MaybeShewill-CV commented 5 years ago

@LukasProgram You may find the node you need in the LaneNet model script:)

LukasProgram commented 5 years ago

@MaybeShewill-CV i think the name of the input node is "input_tensor" and the output node is "lanenet_model" am I right ?

MaybeShewill-CV commented 5 years ago

@LukasProgram No you may print all the variable in the compute graph and select what you need by yourself:)

mbrv commented 5 years ago

@LukasProgram Hi ,

You mentioned your trying to run the code on a smartphone but the reported run-time on NVIDIA 1080 TI is 52 FPS which sounds not very "mobile" friendly. can you elaborate how you plan to use this network on a mobile platform ?

MasschineeMe commented 5 years ago

@mbrv I am trying also to run the network on a mobile phone. After I get the values from the output nodes I need to make the clustering in android. I don't know how the performance will be but let's find out.

Do you have experience with the integration of a network in Android? Do you think it will perform badly?