Polyquin's profile - activity

@berak, I am going through CS231n on youtube. I noticed the slides were saying k nearest neighbours are not good for images. FYI.

Ahh i was trying to cut down on the amount of OCR required, if I can recognise trucks first.

Ah yes I was confused. I'll make the change.

Hi all,

I've been checking a fair bit of Opencv code on the Internet. I noticed most of the code will first preprocess the image (e.g. binarize, apply edges filter) before sending to K nearest for character recognition.

Would it be possible to run K nearest to identify the vehicle (ie. trucks) followed by another K nearest for character recognition. Does it make sense?

Hi all,

I've figured out why the contours decrease when the image resolution is reduced. The reason is due to image interpolation.

If cv2.resize was used, there is a argument for the type of interpolation to be used:

interpolation method:

• INTER_NEAREST - a nearest-neighbor interpolation
• INTER_LINEAR - a bilinear interpolation (used by default)
• INTER_AREA - resampling using pixel area relation. It may be a preferred method for image decimation, as it gives moire’-free results. But when the image is zoomed, it is similar to the INTER_NEAREST method.
• INTER_CUBIC - a bicubic interpolation over 4x4 pixel neighborhood
• INTER_LANCZOS4 - a Lanczos interpolation over 8x8 pixel neighborhood

I tested Hough Lines on this and it detected nothing, so i think it's more than just using findContours now. Any ideas? I'm trying (1) locate the 'rectangle' then recognise the numbers.

Hi all, I am trying to detect the numbers within the white rectangle. My thought process as follows:

1. binarize the image
2. detect canny edges
3. use findContours to detect the contours
4. for each contour, check for numbers

I am stuck at step 3 for this photo. There is a hand cutting the rectangle which lets findContours detect no rectangle. So my question is how to detect and fix an incomplete rectangle.

Ok, I'll show the code and results shortly.

Hi all,

Quick question on findContours. I ran findContours on a picture with resolution of 640 x 960 and findContours returned more contours. I resized the same picture to 320x480 and it returned lesser contours.

I'm not sure why is this happening. I would have thought that the resolution is immaterial to findContour.

Any ideas?

Hi all,

I've compiled opencv 2.4.8 but getting errors like this. The lib and include directories are found in the Makefile. Stumped!

Mac-mini:imagedetection tas5$
./imagedetection  dyld: Library not loaded: lib/libopencv_core.2.4.dylib  
Referenced from:
/Users/tas5/working/opencv_projs/imagedetection/./imagedetection   Reason: image not found Trace/BPT
trap: 5

Makefile

CC = g++ 
CFLAGS = 
LDFLAGS = -I/Users/tas5/local/tbb42/include -I/Users/tas5/local/include -I/Users/tas5/local/include/opencv -I/Users/tas5/local/include/opencv2 -I/Users/tas5/local/include/opencv2/core -L/Users/tas5/local/lib -lm -lopencv_core -lopencv_highgui -lopencv_imgproc -lopencv_video -lopencv_objdetect

ALL = imagedetection all: $(ALL)

imagedetection: imagedetection.o
          $(CC) $(LDFLAGS) -o $@ $^

imagedetection.o: imagedetection.cpp
         $(CC) $(LDFLAGS) -c $<

.PHONY: clean

clean:
         rm -rf *.o core* $(ALL)

Is -w and -h in pixels?

Hey StevenPuttermans, I have a similar question - but I'm detecting faces. I have a bunch of photographs taken and stumped by the -w and -h requirements. These photographs have varying sizes of faces due to the distance between the camera and face and also lens effect. How should I look into fitting in the ratio for -w and -h? Question is, do i need to copy and paste each face individually? what about the ratio if some faces are bigger than others?