CODEWITHSUNDEEP
pythonopencvimage-processingmachine-learning
Pavan R
Pavan R

Reputation: 119

Not able to train a Linear SVM machine

I'm building a SVM linear machine for my image processing project where I'm extracting the features of positive and negative samples and saving it to a directory. I'm then training SVM with these features but I'm getting an error which I'm unable to debug. Below is my train-classifier.py file to train the classifier - 

from skimage.feature import local_binary_pattern
from sklearn.svm import LinearSVC
from sklearn.linear_model import LogisticRegression
from sklearn.externals import joblib
import argparse as ap
import glob
import os
from config import *

if __name__ == "__main__":
    # Parse the command line arguments
    parser = ap.ArgumentParser()
    parser.add_argument('-p', "--posfeat", help="Path to the positive features directory", required=True)
    parser.add_argument('-n', "--negfeat", help="Path to the negative features directory", required=True)
    parser.add_argument('-c', "--classifier", help="Classifier to be used", default="LIN_SVM")
    args = vars(parser.parse_args())

    pos_feat_path =  args["posfeat"]
    neg_feat_path = args["negfeat"]

    # Classifiers supported
    clf_type = args['classifier']

    fds = []
    labels = []
    # Load the positive features
    for feat_path in glob.glob(os.path.join(pos_feat_path,"*.feat")):
        fd = joblib.load(feat_path)
        fds.append(fd)
        labels.append(1)

    # Load the negative features
    for feat_path in glob.glob(os.path.join(neg_feat_path,"*.feat")):
        fd = joblib.load(feat_path)
        fds.append(fd)
        labels.append(0)

    if clf_type is "LIN_SVM":
        clf = LinearSVC()
        print "Training a Linear SVM Classifier"
        clf.fit(fds, labels)
        # If feature directories don't exist, create them
        if not os.path.isdir(os.path.split(model_path)[0]):
            os.makedirs(os.path.split(model_path)[0])
        joblib.dump(clf, model_path)
        print "Classifier saved to {}".format(model_path)

I'm getting an error in the line clf.fit(fds, labels) and below is what it says - 

Calculating the descriptors for the positive samples and saving them
Positive features saved in ../data/features/pos
Calculating the descriptors for the negative samples and saving them
Negative features saved in ../data/features/neg
Completed calculating features from training images
Training a Linear SVM Classifier
Traceback (most recent call last):
  File "../object-detector/train-classifier.py", line 42, in <module>
    clf.fit(fds, labels)
  File "/usr/local/lib/python2.7/dist-packages/sklearn/svm/classes.py", line 200, in fit
    dtype=np.float64, order="C")
  File "/usr/local/lib/python2.7/dist-packages/sklearn/utils/validation.py", line 444, in check_X_y
    ensure_min_features)
  File "/usr/local/lib/python2.7/dist-packages/sklearn/utils/validation.py", line 344, in check_array
    array = np.array(array, dtype=dtype, order=order, copy=copy)
ValueError: setting an array element with a sequence.
Traceback (most recent call last):
  File "../object-detector/test-classifier.py", line 68, in <module>
    fd = hog(im_window, orientations, pixels_per_cell, cells_per_block, visualize, normalize)
  File "/usr/lib/python2.7/dist-packages/skimage/feature/_hog.py", line 63, in hog
    raise ValueError("Currently only supports grey-level images")
ValueError: Currently only supports grey-level images

Upvotes: 0

Views: 531

Answers (2)

susukacang
susukacang

Reputation: 11

I presume the code originated from https://github.com/bikz05/object-detector. You need to make sure that the training samples (pos and neg) have the same size (widthxheight) and are gray images. Your test image should be gray as well.

I use imagemagick's convert command for this:

convert sample.png -resize 100x40 -colorspace gray sample.png

Update (using python to convert to gray image and resize):

import cv2

img = cv2.imread('color_image.jpg',0)
im = cv2.resize(img, (100,40), interpolation=cv2.INTER_CUBIC)
cv2.imwrite("gray_image.jpg", im)

Upvotes: 1

Ekrem Doğan
Ekrem Doğan

Reputation: 694

You can use SVM class of OpenCV instead of scikit's. It's easy to use.

import cv2

# prepare your test and train datasets

svm = cv2.SVM()
svm.train(some_train_data, responses, params)

exp = svm.predict(some_test_data)

For more info, check OpenCV docs.

Upvotes: 0

Related Questions