ndrplz/self-driving-car

Udacity Self-Driving Car Engineer Nanodegree projects.

self-driving-car's Issues

project_12_road_segmentation -- RuntimeError: MetaGraphDef associated with tags 'vgg16' could not be found in SavedModel

Hello,

When I run project_12_road_segmentation with python main.py, I can't get the expected results.
It prompts as

Traceback (most recent call last):
  File "main.py", line 225, in <module>
    run()
  File "main.py", line 176, in run
    image_input, keep_prob, vgg_layer3_out, vgg_layer4_out, vgg_layer7_out = load_vgg(sess, vgg_path)
  File "main.py", line 39, in load_vgg
    tf.saved_model.loader.load(sess, ['vgg16'], vgg_path)
  File "/Users/xxx/.pyenv/versions/3.6.6/lib/python3.6/site-packages/tensorflow/python/saved_model/loader_impl.py", line 209, in load
    "[]") + " could not be found in SavedModel")
RuntimeError: MetaGraphDef associated with tags 'vgg16' could not be found in SavedModel

What is vgg16? How can I get it?
Thank you so much.

SystemError: Unknown opcode

Try to run project 3 on ubuntu18: python3 drive.py got error: SystemError: Unknown opcode

seems it is python version mismatch, ubuntu 18 is python3.6, what version of python do you use to train the model?

Thanks

project_7_unscented_kalman_filter/data/obj_pose-laser-radar-synthetic---What is the last six columns of data in the input file?

self-driving-car/project_7_unscented_kalman_filter/src/main.cpp

Line 118 in 1eadaca

// read ground truth data to compare later

How to run project 3 with my own images and control

I would like to run Project 3 - Behavioral Cloning but with my own images and control. How can I do it? It was not very clear which is the training steep, and which is the inference step. Also, what is the best way to replace the data with my own images and control? My control data also have throttle data, but I can start with just steering.

Thanks for sharing!

Man!Appreciate the sharing!Thanks a lot!
My email : [email protected]
Looking forward more communication!

Webcam image processing

Hello,

I've been trying to implement your code with a webcam I have so I can use it to create an autonomous vehicle that would be able to travel between two white lines. The only issue I have is being able to capture frames from a live camera feed and have the coding process the images. I wouldn't think it would be much harder to do since you have implemented the use of images and video, but I haven't been able to figure out how to do this yet. Also, this will be done from a laptop so no Raspberry Pi will be involved. Any direction or input would be helpful.

Hi. Did you try implment this project by C++?

We know its easier to evaluate or simulate in the environment of Python, but I feel it can be also rewritten and implemented by C++.

from computer_vision_utils.stitching import stitch_together

i can't find computer_vision_utils.stitching ! please help me

project_5_vehicle_detection -- Got TypeError: 'module' object is not callable when run main_ssd.py

When I run main_ssd.py I got this:

Traceback (most recent call last):
  File "/home/binli/self-driving-car/project_5_vehicle_detection/main_ssd.py", line 10, in <module>
    ssd_model, bbox_helper, color_palette = get_SSD_model()
  File "/home/binli/self-driving-car/project_5_vehicle_detection/SSD.py", line 805, in get_SSD_model
    model_ssd = SSD300(input_shape=(300, 300, 3), num_classes=NUM_CLASSES, pretrained=True)
  File "/home/binli/self-driving-car/project_5_vehicle_detection/SSD.py", line 679, in SSD300
    mode='concat', concat_axis=1, name='mbox_loc')
TypeError: 'module' object is not callable

I tried a lot but I cannot fix this? Any suggestions about this? @ndrplz thanks a lot

How to make sure the loop time is 20ms in project_11?

Hello, I cannot find the exact detail to make sure we have a fix-time loop when computing future road points in Project 11 (Path Planning). Could you please tell me where the detail is. Thank you!

CarND-Controls-PID

How to run the carnd-controls-pid program?Can you give me the software and complete instructions?

Distinguishing between dashed and solid lanes

In project_4_advanced_lane_finding, it shows the blue line is for the dashed lane and the red line is for the solid lane. Is the code able to distinguish between the dashed and solid street lanes ?????

How to run inference of capstone_traffic_light_classifier/

For capstone_traffic_light_classifier project, after training, how to run the inference with the test image under "img"?

project_4 : unsupported pickle protocol: %d" % proto

Hello,

I'm stuck in this issue when I run project 4:

$ python ./main.py
Loading cached camera calibration... Traceback (most recent call last):
File "./main.py", line 135, in
ret, mtx, dist, rvecs, tvecs = calibrate_camera(calib_images_dir='camera_cal')
File "/home/nxnam/DigitalRace/hanson/self-driving-car/project_4_advanced_lane_finding/calibration_utils.py", line 20, in wrapper
calibration = pickle.load(dump_file)
File "/usr/lib/python2.7/pickle.py", line 1384, in load
return Unpickler(file).load()
File "/usr/lib/python2.7/pickle.py", line 864, in load
dispatchkey
File "/usr/lib/python2.7/pickle.py", line 892, in load_proto
raise ValueError, "unsupported pickle protocol: %d" % proto
ValueError: unsupported pickle protocol: 3

How can I fix this?

Thank you.

Tensorflow version of SCNN is available now (lane detection)

I have implemented SCNN using Tensorflow and put the full codes here. You can test the code in popular lane detection benchmarks like TuSimple, CULane and BDD100K or your custom dataset with minor modification. Welcome to raising issues if you have problems in reproducing the results. My code is based on LaneNet and SCNN-Torch.

Lane / road marking / vanishing point dataset

VPGNet dataset is available.
It contains around 21K images with human-annotated lane, road markings, and vanishing point labels.
For more details, please check our project page.

how to work by c++ ?

Question 1: Data augmentation

Hello sir,
Here shouldn't be X_train instead of X_train_norm inside train_test_split()?
X_train_norm, X_val_norm, y_train, y_val = train_test_split(X_train_norm, y_train, test_size=VAL_RATIO, random_state=0)
Anyway, here img_rgb = X_train[0], you used X_train[0] which will always give me the first picture of the training dataset! will never be random.
And here: plt.title('Example of RGB image (class = {})'.format(y_train[0])), you used the new training labels (after splitation).

Not sure if you got my point, but it is not correct, I am new to python so I couldn't figure out how to fix the issue here, so any help?

Error in Traffic sign classifier notebook code block 80

project_4 VideoClip actually generates RGB image

In project 4, VideoClip generates RGB image instead of BGR, which would take serious effect on extracting yellow color for thresh_frame_in_HSV() in binarization_utils.py.

Advance Lane Detection - TypeError: 'NoneType' object is not subscriptable

def draw_back_onto_the_road(img_undistorted, Minv, line_lt, line_rt, keep_state):
    height, width, _ = img_undistorted.shape

    left_fit = line_lt.average_fit if keep_state else line_lt.last_fit_pixel
    right_fit = line_rt.average_fit if keep_state else line_rt.last_fit_pixel
    left_fit.sort()
    right_fit.sort()
    # Generate x and y values for plotting
    ploty = np.linspace(0, height - 1, height)
    left_fitx = left_fit[0] * ploty ** 2 + left_fit[1] * ploty + left_fit[2]
    right_fitx = right_fit[0] * ploty ** 2 + right_fit[1] * ploty + right_fit[2]

Error

left_fitx = left_fit_pixel[0] * ploty ** 2 + left_fit_pixel[1] * ploty + left_fit_pixel[2]
TypeError: 'NoneType' object is not subscriptable

Someone please help me with this

Self-driving-Carl

Absence of datasets in capstone traffic light classifier

I am trying to run all the components of the repository but quite a few of them don't have a dataset

Where is the VGG pretrained model in Project 12

I got this error when I try to run the project 12

Traceback (most recent call last):
File "main_27.py", line 284, in
run()
File "main_27.py", line 257, in run
image_input, keep_prob, vgg_layer3_out, vgg_layer4_out, vgg_layer7_out = load_vgg(sess, vgg_path)
File "main_27.py", line 132, in load_vgg
tf.saved_model.loader.load(sess, ['vgg16'], vgg_path)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/saved_model/loader_impl.py", line 200, in load
saved_model = _parse_saved_model(export_dir)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/saved_model/loader_impl.py", line 78, in _parse_saved_model
constants.SAVED_MODEL_FILENAME_PB))
IOError: SavedModel file does not exist at: /home/jack/code/self-driving-car/project_12_road_segmentation/data/vgg/{saved_model.pbtxt|saved_model.pb}
`

Following datas are needed.

You have used train.p and test.p. But I did not find your repository.
Please, give these datasets.

# Load pickled data
train, test = load_traffic_sign_data('../traffic_signs_data/train.p', '../traffic_signs_data/test.p')

Advanced Lane Finding Lane Offset

I have worked for hours and hours with your code. It is wonderful! Thanks for sharing it. However, one thing that makes it hard to implement in an actual bot is the fact that the road curvature and offset values are always positive. Could you please help me make them negative if the road curves, for example, left, and positive if it curves right? Same for the offset within the lane. My self-driving motercycle needs to know which way it is drifting.
Thanks so much!

Error in the 8th Input block of the python notebook

Please refer to this link:link
Error in file:link
Thank you for your time.

Under Lane-changing situations

Hi, Thank you for sharing this wonderful repo!

Just one question: will the lane-detection work the same when the vehicle is lane-changing? how can the left and right lane be detected?

Also, if the road's lane mark is a little bit unclear, will this framework still works?

Trained model

Do you have some trained model for project 12 I can use directly?

project 4

i m unable to access video.
error: from moviepy.editor import VideoFileClip
ModuleNotFoundError: No module named 'moviepy'

i have installed moviepy package but its still not working

ENet-Label-Torch is available now (a light-weight and effective lane detection model)

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

After using the command cmake -G "Unix Makefiles" && make in the build folder

After using the command cmake -G "Unix Makefiles" && make in the build path at the cmd, I have the following error. I have CMake, Make installed and g++ with MinGW
Error:

CMake Error: CMake was unable to find a build program corresponding to "Unix Makefiles". CMAKE_MAKE_PROGRAM is not set. You probably need to select a different build tool.
CMake Error: CMAKE_C_COMPILER not set, after EnableLanguage
CMake Error: CMAKE_CXX_COMPILER not set, after EnableLanguage
-- Configuring incomplete, errors occurred!