||Segmentation and Recognition
Using Structure from Motion Point Clouds, ECCV 2008
Brostow, Shotton, Fauqueur, Cipolla (bibtex)
||Semantic Object Classes in
Video: A High-Definition Ground Truth Database
Pattern Recognition Letters (pdf)
Brostow, Fauqueur, Cipolla (bibtex)
||The Cambridge-driving Labeled
Video Database (CamVid) is the first collection of videos
class semantic labels, complete with metadata. The database
ground truth labels that associate each pixel with one of 32 semantic classes.
Our 32 labels cover most of each image, but our own experiments were normally done with the some of the classes grouped together in a hierarchy. See the "CamToyota Class hierarchy" in Section IV of Classes.txt.
The database addresses the need for experimental data to quantitatively evaluate emerging algorithms. While most videos are filmed with fixed-position CCTV-style cameras, our data was captured from the perspective of a driving automobile. The driving scenario increases the number and heterogeneity of the observed object
Over ten minutes of high quality 30Hz footage is being provided, with corresponding semantically labeled images at 1Hz and in part, 15Hz. The CamVid Database offers four contributions that are relevant to object analysis researchers. First, the per-pixel semantic segmentation of over 700 images was specified manually, and was then inspected and confirmed by a second person for accuracy. Second, the high-quality and large resolution color video images in the database represent valuable extended duration digitized footage to those interested in driving scenarios or ego-motion. Third, we filmed calibration sequences for the camera color response and intrinsics, and computed a 3D camera pose for each frame in the sequences. Finally, in support of expanding this or other databases, we offer custom-made labeling software for assisting users who wish to paint precise class-labels for other images and videos. We evaluated the relevance of the database by measuring the performance of an algorithm from each of three distinct domains: multi-class object recognition, pedestrian detection, and label propagation.
Avi, 30 Mb, xVid compressed. (playback tips or get the free Mac/Windows player.
Mpg, 11 Mb, mpeg-1 compressed (more compatible, but lower quality)
(just samples shown. For all the videos, see below)
||Link to FTP
video files (very big!)
Link to codecs + utility for extracting frames from those big files (read the inventory.txt)
This folder should meet most of your needs, but let me know if not. In brief: Follow instructions in inventory.txt download the big files from the FTP server, install the Panasonic P2 driver for your OS, and run our MXF extractor, feeding it the right script. For the 01TP sequence, you'll also need the Lagarith lossless codec.
(701 so far)
Link to zip file with painted class labels for stills from the video sequences.
Txt file listing classes and label colors as RGB triples (sorted).
(Note: the corresponding raw input images only - at 1Hz,
already extracted from the respective videos are here (556Mb).)
to files and code
The relevant line that you care about to get the projection matrix of 1 camera is in MotBoostEvalOneFrame.m (see how LoadBoujou_2Dtrax_3dBans_Misc.m calls it):
curC = Cs( frameNum-offsetForFrameNums, 1:3);
stored .mat files in this Data
folder. Note, the main file is
EgoBoost_trax_matFiles.zip, and Seq01TP's .mat files are
there too, but in the separate 01TP__AllTrax.zip.
Example camera pose trajectory, stored in Boujou Animation Format:
each line containing "AddDecompCameraKey" has a K and R matrix and t vector,
so that P = K * R * [I -t]
Description: 6120 frames at 30Hz == 3:24 min
VideoFiles 1 and 2 in MXF format* (note: these are 2 halves of 1 zip file)
(see also CamSeq01)
Description: 5130 frames at 30Hz == 2:51 min
VideoFile in MXF format*
|Listing of (RGB)-Class assignments (alphabetical) Listing in color-order used by MSRC (with "XX")|