Fig. 7
From: A novel method for robust markerless tracking of rodent paws in 3D
The error plots. Panels a, b, and c show a comparison between our proposed algorithm with four other methods for the dataset (4 cameras × 2 trials each animal × 5 mice = 40 trials). These three methods are SLIC with 2D KF, “SLIC+2D”; SLIC with 3D KF, “SLIC+3D;” and MT. Panel a shows the required time by different methods to process one trial, 1000 frames from each of the four cameras. Panels b and c show the average number of “critical mistakes” per trials that software needed user intervention to label paws correctly after that mistake (major error) and the average number of mistakes per trials that the software did not need any user intervention to recover from the mistake (minor error), respectively. While panels d, e, and f show the results of each trial for the measures presented in panels a, b, and c, respectively. The colors show the results for the methods presented in a, b, and c. 3D reconstruction error, Section 4.4, is shown in g. It shows the error for the calibration object, front paw, and hind paw in the four cameras. The blue, red, yellow, and green colors in g show the results of camera 1, 2, 3, and 4. The plots h and i show the 3D reconstruction error of different trials for front and hind paws, respectively