awful OpenCV Haar Cascades code

haar_cascades/OpenCV_haar_cascades.md

@@ -0,0 +1,21 @@
+## Instructions
+
+First off, the OpenCV Haar Cascades training feature is awful. There has to be a better way. Anyways, here are steps.
+
+	python negative_images.py
+	cd /Users/ryanzotti/Documents/repos/opencv-haar-classifier-training
+	find ./positive_images -iname "*.jpg" > positives.txt
+	find ./negative_images -iname "*.jpg" > negatives.txt
+	perl bin/createsamples.pl positives.txt negatives.txt samples 1500 "opencv_createsamples -bgcolor 0 -bgthresh 0 -maxxangle 1.1 -maxyangle 1.1 maxzangle 0.5 -maxidev 40 -w 80 -h 40"
+	conda create --name python2 python=2
+	source activate python2
+	opencv_traincascade -data classifier -vec samples.vec -bg negatives.txt -numStages 20 -minHitRate 0.999 -maxFalseAlarmRate 0.5 -numPos 1000 -numNeg 600 -w 80 -h 40 -mode ALL -precalcValBufSize 1024 -precalcIdxBufSize 1024
+
+## References
+
+* http://coding-robin.de/2013/07/22/train-your-own-opencv-haar-classifier.html
+* https://github.com/mrnugget/opencv-haar-classifier-training
+* http://www.trevorsherrard.com/Haar_training.html
+* http://docs.opencv.org/3.1.0/dc/d88/tutorial_traincascade.html
+* http://www.pyimagesearch.com/2016/06/20/detecting-cats-in-images-with-opencv/
+* http://note.sonots.com/SciSoftware/haartraining.html

haar_cascades/negative_images.py

@@ -0,0 +1,20 @@
+import cv2
+import numpy as np
+from random import randint
+
+negative_image_file_path = "/Users/ryanzotti/Documents/repos/Self_Driving_RC_Car/final_processed_data_3_channels.npz"
+output_path = "/Users/ryanzotti/Documents/repos/opencv-haar-classifier-training/negative_images"
+npzfile = np.load(negative_image_file_path)
+train_predictors = npzfile['train_predictors']
+shape = train_predictors.shape
+total_frames = shape[0]
+height, width = shape[1], shape[2]
+negative_images = []
+negative_images_count = 3000
+for i in range(negative_images_count):
+    random_index = randint(0, total_frames)
+    random_image = train_predictors[random_index]
+    #cv2.imshow(str(random_index),random_image)
+    random_image_file_name = output_path+"/random_image_"+str(i)+".jpg"
+    cv2.imwrite(random_image_file_name, random_image)
+print("Finished")

show_stop_signs.py

@@ -0,0 +1,62 @@
+import cv2
+import sys
+from util import shell_command
+import numpy as np
+from random import randint
+
+
+def process_image_set(dir,pixel_size):
+    processed_images = []
+    for file_name in file_names:
+        image = cv2.imread(dir + "/" + file_name)
+        height, width = image.shape[0], image.shape[1]
+        if width == height:
+            resized = cv2.resize(image, (pixel_size, pixel_size), interpolation=cv2.INTER_AREA)
+            processed_images.append(resized)
+        else:
+            print(file_name + " is not a square! It has dimensions " + str(image.shape))
+            sys.exit(1)
+    processed_images = np.array(processed_images)
+    return processed_images
+
+# Generates negative images from previous driving session.
+def random_negative_images():
+    pass
+
+pixel_size = 24
+
+# Assumes positive images are each separate files
+positive_image_dir = "/Users/ryanzotti/Dropbox/StopSigns"
+shell_cmd = 'ls {dir}'.format(dir=positive_image_dir)
+dir_contents = str(shell_command(shell_cmd)).replace("b\'","").split("\\n")
+file_names = [file_name for file_name in dir_contents if "JPG" in file_name]
+positive_images = process_image_set(positive_image_dir,pixel_size)
+
+# Assumes negative images are taken from existing numpy zip file.
+# I used my training data. This is ideal because it is
+# representive of backgrounds that I would realistically see in a
+# production setting
+negative_image_file_path = "/Users/ryanzotti/Documents/repos/Self_Driving_RC_Car/final_processed_data_3_channels.npz"
+npzfile = np.load(negative_image_file_path)
+train_predictors = npzfile['train_predictors']
+shape = train_predictors.shape
+total_frames = shape[0]
+height, width = shape[1], shape[2]
+negative_images = []
+negative_images_count = 3000
+for i in range(negative_images_count):
+    random_index = randint(0, total_frames)
+    random_image = train_predictors[random_index]
+    cv2.imshow(str(random_index),random_image)
+    height_bound = height - 24
+    width_bound = width - 24
+    random_height_start = randint(0,height_bound)
+    random_height_end = random_height_start + pixel_size
+    random_width_start = randint(0,width_bound)
+    random_width_end = random_width_start + pixel_size
+    random_image_section = random_image[random_height_start:random_height_end,random_width_start:random_width_end]
+    negative_images.append(random_image_section)
+negative_images = np.array(negative_images)
+#np.savez(session_path + '/predictors_and_targets', predictors=predictors,targets=targets)
+
+print("Finished")