When I run conda env create -f lane_environment.yml I get the following error: Collecting actionlib==1.11.7 Could not find a version that satisfies the requirement actionlib==1.11.7

If actionlib is required for this, can you maybe update the environment file with another compatible version?

Thanks

Hi Michael. Thank you so much for this intensive work! I wonder in your final method, is there any way that could calculate the offset of the car from lanes, comparing with the method that uses perspective transformation?

Hello. First, cong. Awesome project, I was wanting to learn deep learning with making lane detection. I can learn now but how can I control the motors sir? How we can calculate the lanes? I have idea with using Tan90=Sin90/Cos90 like that.

Have a nice day!

Hi, Iam facing following issues:

1.I am not able to open the full_CNN_train.p and CNN_label.p in Matlab. 2.How can I train and label my own dataset using this code?

Please can you guide where I am doing wrong

I tried to run this lane_lines.py and got the following error:

Traceback (most recent call last): File "lane_lines.py", line 26, in label_scaler = pickle.load(open( "scaler.p", "rb" )) FileNotFoundError: [Errno 2] No such file or directory: 'scaler.p'

Dear @mvirgo , Thanks for creating such an amazing project! I am sure that I will learn from it. For now, I am using this as a benchmark for my new PC and I was wandering where the file 'Full_CNN_train.p' is as it is giving me a File Not Found error.

I simply cloned the github repo. by .p, do you mean .py? Many thanks 👍

Clement Secondary School Student :)

when i run the check_labels.py in line 36 the issue is ValueError: operands could not be broadcast together with shapes and one more question what is list index out of range. Can you help me pls?

I use this code to get output file which is a road with detected path according to detected lines:

from os.path import splitext, basename

import numpy as np import cv2

from moviepy.editor import VideoFileClip from keras.models import load_model

model = load_model('full_CNN_model.h5')

class Lanes(): def init(self): self.recent_fit = [] self.avg_fit = []

lanes = Lanes() vid_in = "line_1.mp4" vid_output = 'line_1_out.mp4'

video_capture = cv2.VideoCapture() fourcc = cv2.VideoWriter_fourcc(*'XVID') name, extension = splitext(basename(vid_in)) output = name + '.avi'

if video_capture.open(vid_in): property_id = int(cv2.CAP_PROP_FRAME_COUNT) total_frames = int(cv2.VideoCapture.get(video_capture, property_id)) frame_no = 1 width, height = int(video_capture.get(cv2.CAP_PROP_FRAME_WIDTH)),
int(video_capture.get(cv2.CAP_PROP_FRAME_HEIGHT)) fps = video_capture.get(cv2.CAP_PROP_FPS)

video_writer = cv2.VideoWriter(output, fourcc, fps, (width, height))
while video_capture.isOpened():
    ret, frame = video_capture.read()
    if not ret:
        break
    
    small_img = np.array(np.resize(frame, (80, 160, 3)))
    small_img = small_img[None,:,:,:]
    
    prediction = model.predict(small_img)[0] * 255

    lanes.recent_fit.append(prediction)
    if len(lanes.recent_fit) > 5:
        lanes.recent_fit = lanes.recent_fit[1:]

    lanes.avg_fit = np.mean(np.array([i for i in lanes.recent_fit]), axis = 0)

    blanks = np.zeros_like(lanes.avg_fit).astype(np.uint8)
    lane_drawn = np.dstack((blanks, lanes.avg_fit, blanks))
    lane_image = np.resize(lane_drawn, (720, 1280, 3))

    result = cv2.addWeighted(lane_image, 1, lane_image, 1, 0)


    if frame_no % 20 == 0:
        print('# {}/{} frames processed!'.format(frame_no, total_frames))

    frame_no += 1
    video_writer.write(frame)
video_capture.release()
video_writer.release()

I also used the code in the original repository, but I did not get the same output in repository, Although I use the same steps.

Do I have to train another model now?

Please post the training loss graph and validation loss graph for your trained model. I am using as a reference for my dissertation. It will be very helpful.

how to make labels without that redrawn lanes ? or how to get those redrawn lanes can we change the final output to be a convolution and not a deconv and what should the output be

I got an error after i had run python3 MLND-Capstone/draw_detected_lanes.py:

Traceback (most recent call last): File "MLND-Capstone/draw_detected_lanes.py", line 9, in model = load_model('full_CNN_model.h5') File "/usr/lib/python3.7/site-packages/keras/engine/saving.py", line 458, in load_wrapper return load_function(*args, **kwargs) File "/usr/lib/python3.7/site-packages/keras/engine/saving.py", line 549, in load_model with H5Dict(filepath, mode='r') as h5dict: File "/usr/lib/python3.7/site-packages/keras/utils/io_utils.py", line 191, in init self.data = h5py.File(path, mode=mode) File "/usr/lib/python3.7/site-packages/h5py/_hl/files.py", line 394, in init swmr=swmr) File "/usr/lib/python3.7/site-packages/h5py/_hl/files.py", line 170, in make_fid fid = h5f.open(name, flags, fapl=fapl) File "h5py/_objects.pyx", line 54, in h5py._objects.with_phil.wrapper File "h5py/_objects.pyx", line 55, in h5py._objects.with_phil.wrapper File "h5py/h5f.pyx", line 85, in h5py.h5f.open OSError: Unable to open file (unable to open file: name = 'full_CNN_model.h5', errno = 2, error message = 'No such file or directory', flags = 0, o_flags = 0)

1# I got an error of different input types on add weighted method so I've to specify an output type

before change: result = cv2.addWeighted(image, 1, lane_image, 1, 0) after change: result = cv2.addWeighted(image, 1, lane_image, 1, 0, dtype= cv2.CV_32F).astype(np.uint8)

now there are no lanes on mn output video 2# and also I've used "from skimage.transform import resize as imresize" for image resize but I didn't change any other lines related to imresize is it ok? i.e lane_image = imresize(lane_drawn, (720, 1280, 3)) its same before and after

Hello, I'm not understanding how you are creating labels and what's inside in labels.p file as you are dividing labels by 255.Why?

Normalize labels - training images get normalized to start in the network

labels = labels / 255 Please guide me in this. Thank You!

The current way the dataset is loaded for training is super inefficient and loads the whole dataset all at once. As such, I should consider changing the dataset from being stored as a pickle file, as well as whether to use flow_from_directory or similar techniques.

Scipy deprecated imresize as of 1.3.0 and now suggests using PIL to resize images. This will require both updating code as well as the environment file to eventually remove scipy and replace it with pillow.

Can also consider using alternative image re-sizing functions from other libraries as well.