- Ryota Yoshihashi, Rei Kawakami, Makoto Iida, and Takeshi Naemura: ``Bird Detection and Species Classification with Time-lapse Images Around a Wind Farm: Dataset Construction and Evaluation, Wind Energy 20.12 (2017): 1983-1995.
- Ryota Yoshihashi, Rei Kawakami, Makoto Iida, and Takeshi Naemura: ``Evaluation of Bird Detection using Time-lapse Images around a Wind Farm,'' European Wind Energy Association Conference (EWEA2015) (2015.11).
- Ryota Yoshihashi, Rei Kawakami, Makoto Iida, and Takeshi Naemura: ``Construction of a Bird Image Dataset for Ecological Investigations,'' IEEE Intern. Conf. Image Process. (ICIP2015), pp. 4248 -- 4252 (2015.9).
The annotation txt file have the following format:
image file name 1 x1,y1,w1,h1,(b, n, or u) x2,y2,w2,h2,(b, n, or u) ... x_n,y_n,w_n,h_n,(b, n, or u) image file name 2 ... image file name n ...
In this file, "image file name n" means the name of image file. Following comma-separated lines are values describe the bounding boxes. The line of "x1,y1,w1,h1,(b, n, or u)" means a bounding box whose left-top point is (x1, y1), width is w1 and height is h1. The last character element represents the class of the bounding box: "b" means birds, "n" non-birds and "u" undefined objects. Read it in your program and you can use the labels for each bounding box.