search

google-research-datasets / Objectron

Objectron is a dataset of short, object-centric video clips. In addition, the videos also contain AR session metadata including camera poses, sparse point-clouds and planes. In each video, the camera moves around and above the object and captures it from different views. Each object is annotated with a 3D bounding box. The 3D bounding box describes the object’s position, orientation, and dimensions. The dataset contains about 15K annotated video clips and 4M annotated images in the following categories: bikes, books, bottles, cameras, cereal boxes, chairs, cups, laptops, and shoes
Other
2.24k stars 263 forks source link

Possible bug in from_transformation in box.py #59

Open melecmat opened 2 years ago

melecmat commented 2 years ago

Hi, I noticed that in the class method from_transformation in box.py, the transformation itself is thrown away, only the points are kept, upon calling cls(vertices=vertices). That cannot be right? For example, in IOU computation, the boxes are transformed by these transforms, so when they are not set it leads to different results.

Best, Matous

ahmadyan commented 2 years ago

it is working as intended. It will keep the vertices, and sets everything else to None in the constructor. Every time if you access any of the transformation, rotation, translation, etc. if it is None it will fit a transformation to the vertices via [.fit()]( def translation(self):).

melecmat commented 2 years ago

Ah, OK, now I see it. Thanks.

Out of curiosity, can I ask what is the main reason for this design choice?