mitmul/caltech-pedestrian-dataset-converter

Download Caltech Pedestrian Dataset and convert them for Python users without using MATLAB

caltech-pedestrian-dataset-converter's Issues

openCV version and cv2.VideoWriter_fourcc()

There's a minor difference in openCV version 2.x and 3.x, it might be nice to use something like this:

if opencv_version() == 2:
	fourcc = cv2.cv.FOURCC(*output_format)
elif opencv_version() == 3:
	fourcc = cv2.VideoWriter_fourcc(*output_format)
else:
	raise

[mjpeg @ 0x55e5f56bbc40] No JPEG data found in image

hello， when i convert seq to image，i meet [mjpeg @ 0x55e5f56bbc40] No JPEG data found in image。。

Documentation on the converted files

Hi! It is easy to run the scripts and get the files, but how can we use the converted files, especially the annotations, eg, if we read the annotations.json, then under each sequence, there is filed of [u'altered', u'log', u'nFrame', u'logLen', u'frames', u'maxObj'], what does those means? how to use the converted annotations? could you pls give a doc or ref a doc for those? I have checked official caltech website, but still found nothing related, thanks!

How can I get Caltech reasonable dataset???

Caltech is divided to overall,scale.occlusion,reasonable. I can just get the overall version. So how to get the reasonable part??? @mitmul

python scripts/convert_seqs.py is not generating images from the caltech dataset

The script didn't throw any errors and print all the video file names but no images were generated in the data/images folder as i was expecting. Would you know what could be wrong?

output images format (PNG vs. JPG)

Nice utilities, saved me maybe a day or two writing similar stuff.

One minor problem, though, the caltech 2014 dataset is basically .jpg files packed in a sequence. Unpacking them as .png consumes probably several times more of disk space, compared to .jpg

You may want to change this in your scripts.

''python scripts/convert_seqs.py'' not working for me

michel@michel-Lenovo-IdeaPad-Y510P:~/Desktop/caltech-pedestrian-dataset-converter-master$ python scripts/convert_seqs.py
GStreamer Plugin: Embedded video playback halted; module typefind reported: Could not determine type of stream.
OpenCV Error: Unspecified error (GStreamer: unable to start pipeline
) in cvCaptureFromCAM_GStreamer, file /home/michel/opencv-2.4.13/modules/highgui/src/cap_gstreamer.cpp, line 816
Traceback (most recent call last):
File "scripts/convert_seqs.py", line 19, in
cap = cv.VideoCapture(fn)
cv2.error: /home/michel/opencv-2.4.13/modules/highgui/src/cap_gstreamer.cpp:816: error: (-2) GStreamer: unable to start pipeline
in function cvCaptureFromCAM_GStreamer

Can this repo convert other dataset in .seq format?

I try use this repo to convert the dataset below, but i encounter the following error.

Unable to stop the stream: Inappropriate ioctl for device
data/set05/V002.seq

GM-ATCI Rear-view pedestrians dataset
https://sites.google.com/site/rearviewpeds1/

Who generated the problem of Image time and space?