talkera.org's profile - activity

To recognize the car model you have to use a supervised machine learning method. The program can decide based on features. You could try template matching if they are simple 2d cars. If you need more advanced features you could look into key point algorithms or statistical features.

While there is no PHP OpenCV support, you can do a trick if you have a unix machine. Simply start the C version of face detection, modify it to save the data. Run the binary from your php web application. This will only work with static images. If you need more, go with Java. (Python could also be an option, it is supported by OpenCV.)

The easiest solution is to use key point detection / matching to match the label. Since the number of bottle brands is limited this is probably the most easy and accurate approach. This is possible if there are some distinct images on the bottle such as a logo of a wine company. The program would match which features of the label or bottle are most similar and find the necessary data in a database.

If you want the computer to read all of the characters on the label the accuracy will be much lower and complexity of the code will be increased. You would have to flatten the label and feed the label image to an ocr engine.

NoClassDefFoundError is a typical Java error. The NoClassDefFoundError is given when there is a class file that your code depends on and it is present at compile time but not found at runtime. Do the example codes from OpenCV work?

A simple solution could be to try a nearest neighbor approach, which may or may not work but will certainly be very slow. For every white pixel in the image find the nearest white pixels within a short distance range. If found, create a line between the two points. This could be speed up by skipping pixels;

An even simpler solution would be be to blur the image a few times, such that you would have a vague gray blur in the gap area. Then threshold the image such that anything not black becomes white.

Get the cascade and put it in the same directory.

wget https://jviolajones.googlecode.com/files/haarcascade_frontalface_default.xml

Now there was a typo in your code: haarScascade_frontalface_default.xml should be haarcascade_frontalface_default.xml

Change your code to this, make sure the indention is correct and it will work (make sure the cascade xml file is in the same directory):

import numpy as np
import cv2

face_cascade = cv2.CascadeClassifier('haarcascade_frontalface_default.xml')
video_capture = cv2.VideoCapture(0)

while True:
    ret, frame = video_capture.read()
    gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)

    faces = face_cascade.detectMultiScale(gray,1.3,5)
    for(x,y,w,h) in faces:
       cv2.rectangle(frame,(x,y),(x+w,y+h),(255,0,0),2)  
       print str(x)

   cv2.imshow('video', frame)
   if cv2.waitKey(1) & 0xFF == ord('q'):
     break
 video_capture.release()
 cv2.destroyAllWindows()

Tested with Python 2.7.6. This works on my machine. If that still does not work copy the code with static image face detection from here and see if that works.

You can access color data, but color itself does not tell anything about the structure of the objects. Did you try various values for the hessian matrix? Is your training set any similar to the test set? You could try HOG features or statistical features. Adding images would be useful.

The algorithm computation time may differ from the paper because of the implementation. In OpenCV the SURF algorithm should be faster than SIFT, but at the time that I tested it the implementation of SIFT was faster. A big impact on the computation time is the implementation of the algorithm. This implementation of SURF has another computation time. So what can you do?

You could try to enable hardware acceleration or to create a hardware accelerated implementation yourself. The OpenCV library has GPU support. Here is an interesting page http://opencv.org/platforms/cuda.html on hardware speed up and an example.

Another thing you could try is to downscale the image. This will result in less key points being found, but they will be found faster.

Finally, have a look at this thread http://www.answers.opencv.org/question/7002/freak-compute-speed-up/

If nothing works, get a faster computer :)

I think you made an accidental small change in the code. Download the original tutorial code again and try to run that. If that does not work anymore, something in the settings is wrong.