error while decoding MB 4 6, bytestream -9
baharna2 opened this issue · comments
Hi,
Thank you so much for providing this repo.
I'm running your code in a virtual machine which I only have access to the terminal, so I decided to save each frame instead of displaying it. I get this error when running the code
error while decoding MB 4 6, bytestream -9
The code stops after processing first frame of video file. I'd be thankful if you kindly help on that.
Thanks,
Bahareh
-
Saving counting result as a new video is very easy task. By the way, I am also planning to add this feature in TensorFlow Object Counting API.
-
"There are some inconsistent indentations for tabs and spaces that arises while running." -> Could you specify those lines.. I already fixed this kind of indentation issues.. It can be caused by your python version..
-
"There are also some errors when importing several modules which you kindly provided." -> It must be caused by your python version, i think you use python 3.x version. This sample project have been developing on python 2.7.x version and "import" is different on python 3.x versions. However, I am also planning to publish python 3.x version of this project soon.
-
"Additionally, the module you provided for the direction does not seem to be general. It can be applied to very specific scenes." -> Yes because this repo is just a sample project to prove the TensorFlow Object Counting API can be used as a base architecture to develop object counting systems. You can find more generalized and lightweight modules for object counting tasks in this repo. You can develop object counting systems for very different types of scenarios using the TensorFlow Object Counting API.
The project has been updated from python2 to python3 and all the indentation problem fixed.
"However there is a more serious problem. It doesn’t detect any of the vehicles in the video, even when I changed the utilized model to faster rcnn, mask rcnn etc, however I tried to save the result for one of the video frame using object detection API github only (not your project code) through ssd mobilenet and it’s able to partially detect the card. There should be something wrong with either vehicle-detection-main or visualization_utils script. Do you have any idea. There is no error, but no car is detected when using your code." -> Could you prove it by any screen shots or any videos? It looks not possible that TenworFlow Object Detection API can detect the some objects that my project can not detect because my project has been built on top of TensorFlow Object Counting API that has been built on top of TensorFlow Object Detection API..