The deliverable is an API written in python that accepts a JSON POST request:
multiple frames of images with bounding box locations.
The system should be able to keep track of similar items between the images. It should also be able to give unknown items a new ID and start tracking them from that point on.
A good example for the system we are looking for a Filter. You are free to use a Filter or any other type of algorithm to keep track of detections between frames.
The return value of the API contains the list of frames, bounding boxes and an ID number per bounding box.
The system should handle occlusion well. Sometimes items are obscured between frames, but appear several frames later.
We are looking for anyone or any group capable of delivering what we are looking for. Experience is nice to have.
POST request data example:
{
'data': [
{
'frame_number': 1,
'base64_image': '{ image data here...} ',
'detection_boxes': [
{
'boxes': [
{
'x1': 0,
'y1': 0,
'x2': 200,
'y2': 250,
'confidence': 0.59
} ,
{
'x1': 500,
'y1': 500,
'x2': 650,
'y2': 600,
'confidence': 0.9
}
]
}
]
}
]
}
RESPONSE example:
{
'data': [
{
'frame_number': 1
'detection_boxes': [
{
'boxes': [
{
'x1': 0,
'y1': 0,
'x2': 200,
'y2': 250,
'confidence': 0.59,
'id': 1
} ,
{
'x1': 500,
'y1': 500,
'x2': 650,
'y2': 600,
'confidence': 0.9,
'id': 2
}
]
}
]
}
]
}
About the recuiterMember since May 20, 2018 Anurag Nigam
from Bihar, India