From a41cb1a4d66890a3cb758900984d389687331c4c Mon Sep 17 00:00:00 2001 From: Yusuke Niitani Date: Fri, 9 Jun 2017 16:08:26 +0900 Subject: [PATCH] fix ambiguity in the explanation of bounding box coordinates convention --- README.md | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/README.md b/README.md index 4903a1d51b..0fd47bedb7 100644 --- a/README.md +++ b/README.md @@ -45,7 +45,8 @@ ChainerCV follows the following conventions. + Channels of a color image are ordered as RGB. The shape of an image-array is CHW (channel, height, width). + Axis-dependent values follow row-column order. + Shape of an image is always represented as `(height, width)`. - + Coordinates of bounding boxes and keypoints are ordered as `(y, x)`. + + Coordinates of bounding boxes are ordered as `(y_min, x_min, y_max, x_max)`. `(y_min, x_min)` and `(y_max, x_max)` are coordinates of the top left and the bottom right vertices of a bounding box respectively. + + Coordinates of keypoints are ordered as `(y, x)`. ## Models Currently, ChainerCV supports networks for object detection and semantic segmentation.