ENPM673-Project1

This project tracks up to 3 AR tags in a video. Examples of the 3 tags (after processing) can be seen below. These tags are divided into an 8x8 grid for orientation analysis and identification.

Program Modes

The program has several modes, all accessible via a series of Boolean toggles at the top of main.py.

Write_to_video

This mode, when set True, saves the output of the video to an AVI file. The video will save in the same directory as main.py. The name of the video is determined by the modes you have enabled, so that the output video filename is a clear log of the configuration it recorded. If you only want to export a video (to watch later) we recommend toggling Fast_mode to True. This disables previewing the frames on screen, and offers some performance increases. Fast_mode is not recommended in conjunction with any other mode.

show_contours

You have the option to enable the display of the contours found in the frame. Green contours denote the outer edge of the paper. The blue contours denote the edge of the AR tag. This mode can be enabled alongside any other mode.

Dog_mode

Dog_mode, when toggled True, takes our 3 dog images and superposes them on their respective tags.

Cube_mode

Cube_mode displays a cube on each tag. You can run Dog_mode and Cube_mode at the same time. If you do, the cubes will render transparent so that the dogs can be seen on the bottom face. If you run just Cube_mode, the cubes will render with grey faces.

You can change the color of the cube faces by editing face_color near the top of main.py. The colors are in (B,G,R) from 0-255. Similarly, you can edit the color of the cube edges by changing edge_color near the top of main.py. The cube faces can also be colored to correspond to their unique tag IDs.

video_src

This allows you to change which of the three available videos are used in the program.

video_src = 1 is the first video (data_1.mp4), which has 1 AR tag.

video_src = 2 is the second video (data_2.mp4), which has 2 AR tags.

video_src = 3 is the third video (data_3.mp4), which has 3 AR tags.

You can use any of these three videos with any combination of other modes.

Smooth Cubes

There is another executable program in this set: smooth_cubes.py. This program renders cubes with less jittery motion. It uses a forward and backward looking approach, averaging the cube corner positions over several preceeding and post frames. Similarly to main.py, there are several customizable program parameters that can be adjusted at the top of smooth_cubes.py:

Once again, write_to_video toggles saving the output video file and video_src changes the input source video. num_future_frames is the number of frames (in both directions) to average the points over. This smoothing function works for an arbitrary number of preceding and post frames, allowing for varying amounts of smoothing. The more frames that are averaged results in a smoother, less jittery cube, but may cause cube drift, where the cube position slides slightly from the tag due to camera movement in neighboring frames. Cube drift becomes more noticeable as the frame count increases.

Notes on Usage

Everything is designed to run in Python 3.7. Download the entire directory before beginning (all files are required). Open main.py and adjust the toggle parameters to the configuration you want to run. Save the file and run python3 main.py in Terminal. The run may take several minutes, depending on the configuration selected.

You can also run smooth_cubes.py in the same fashion. We do not recommend running main.py and smooth_cubes.py simultaneously.