Visual Bounding Box Preprocessing | Dataford Interview Questions

Implement an efficient algorithm to crop, scale, and align visual bounding boxes or video frame sequences for dataset preprocessing.

Implement crop_scale_boxes(frame_boxes, crop, target_size). frame_boxes is a list of frames, where each frame contains boxes [x1, y1, x2, y2] in source-image coordinates. crop is [left, top, right, bottom], and target_size is [width, height]. Intersect every box with the crop, translate it to crop-relative coordinates, scale it to the target size, and discard boxes with zero visible area. Preserve frame order and box order.

Return a list of frames containing transformed boxes. Coordinates may be floating-point values.

Constraints

1 <= len(frame_boxes) <= 500
0 <= len(frame_boxes[i]) <= 500
Each box is [x1, y1, x2, y2] with x1 < x2 and y1 < y2
crop is [left, top, right, bottom] with left < right and top < bottom
target_size is [width, height] with positive dimensions
Coordinates and dimensions are integers between 0 and 10^6

Function Signature

def crop_scale_boxes(frame_boxes, crop, target_size):

Your question is Visual Bounding Box Preprocessing. Start with the requirements in the Question tab.

Run and submit as often as you like. When you're ready, talk me through your approach or go straight to the code.

Problem

Implement an efficient algorithm to crop, scale, and align visual bounding boxes or video frame sequences for dataset preprocessing.

Return a list of frames containing transformed boxes. Coordinates may be floating-point values.

Constraints

1 <= len(frame_boxes) <= 500
0 <= len(frame_boxes[i]) <= 500
Each box is [x1, y1, x2, y2] with x1 < x2 and y1 < y2
crop is [left, top, right, bottom] with left < right and top < bottom
target_size is [width, height] with positive dimensions
Coordinates and dimensions are integers between 0 and 10^6

Function Signature

def crop_scale_boxes(frame_boxes, crop, target_size):

Your question is Visual Bounding Box Preprocessing. Start with the requirements in the Question tab.

Run and submit as often as you like. When you're ready, talk me through your approach or go straight to the code.