Considering object localization problem as a markov decision process and making an agent learn to maximize its reward on it. Implementation based on paper : Caicedo, Juan Lazebnik, Svetlana. (2015).”Active Object Localization with Deep Rein-forcement Learning”. 10.1109/ICCV.2015.286, arxiv
Number of training elements per class in VOC2007 + VOC2012
cat : 648 elements.
bird : 553 elements.
motorbike : 304 elements.
diningtable : 188 elements.
train : 369 elements.
tvmonitor : 290 elements.
bus : 268 elements.
horse : 310 elements.
car : 659 elements.
pottedplant : 202 elements.
person : 1301 elements.
chair : 379 elements.
boat : 289 elements.
bottle : 258 elements.
bicycle : 303 elements.
dog : 750 elements.
aeroplane : 432 elements.
cow : 210 elements.
sheep : 208 elements.
sofa : 297 elements.
