CVC-02 Pedestrian Dataset ------------------------- + Info ---- This is the CVC-02 pedestrian dataset developed in the Computer Vision Center. The dataset is available at www.cvc.uab.es/adas, and has been developed and annotated by David Gerónimo for his PhD Thesis, together with Dr. Angel D. Sappa, who provided invaluable help with the recording of the sequences. + Available data -------------- This dataset contains data in different formats, depending on the information: - color : the original rectified right color image of the stereo pair. (png format) - annotations : the annotated pedestrians in the range from 5 to 50 m in the color image. (txt format) - 3dpoints : reconstructed 3D points of the image. (own format) - depth : the image containing the depth value of each pixel (png format) + Annotations format ------------------ Each .txt (one per image) is a text file that contains the annotations of the image in the following format: x y w h x y w h ... x y w h where (x,y) is the center pixel and (w,h) is the width and height of each window. Here is a C example code to read the files: ---------------------------------------------------------------------------- | |\ | void ReadAnnotations(char* filename) | \ | { | \ | int x, y, w, h; | \ | ----| | FILE* fp = fopen(filename, "r"); | | | | while(fscanf(fp, "%d %d %d %d ", &x, &y, &w, &h) == 4) | | { | | char a[100]; | | fscanf(fp, "%s\n", a); | | | | if (!strcmp(a, "PEDESTRIAN-OBLIGATORY")) | | { | | // create new annotation in your own structure with x, y, w and h | | vAnnotations.push_back( new CAnnotation(x,y,w,h) ); | | } | | } | | fclose(fp); | | } | | | -------------------------------------------------------------------------------- + 3dpoints format --------------- Each .pts (one per image) is a binary file that contains the reconstructed 3D points data of the scene. The format is the following: imgwidth imgheight <- width and height of the image (in fact it is always 640 and 480) u v x y z r g b <- \ u v x y z r g b <- | ... | 2D image (u,v) and 3D scene (x,y,z) coordinates and color (r,g,b) of the points u v x y z r g b <- / where u,v are integers (4bytes), x,y,z are floats (4bytes) and r,g,b are unsigned chars (1 byte). There are width*height rows of points. Here is a C example code to read the files: ---------------------------------------------------------------------------- | |\ | void Read3DPoints(char* filename) | \ | { | \ | int width, height; | \ | ----| | FILE*fp=fopen(filename, "rb"); | | | | fread(&width, sizeof(int), 1, fp); | | fread(&height, sizeof(int), 1, fp); | | | | for (long i=0; i