Missing tracking points of object

Hey all,

I am trying to train a network to identify the location of an infant and the surrounding objects (toys, and other people in the room).
I have marked overall more than 1900 frames across 60 videos. I trained the net until the loss converged.

When I run the command deeplabcut.create_video_with_all_detections, I get a video with everything being tracked (although the tracking of anything besides the infant is not very good). However, when I extract the coordinates to a csv file, only the infants has any coordinates being shown (in the CSV file). That is, the columns describing the coordinates of everything which is not the infant are empty.

Here are the relevant parameters from the config_file:


  • infant
    uniquebodyparts: [parent left leg, parent right leg, camera woman left leg, camera
    woman right leg, red ball, broom, popper, bucket with balls, doll, stroller]
    multianimalbodyparts: [infant head, infant face, infant left leg, infant right leg,
    infant left sholder, infant right sholder]


    • infant head
    • infant face
    • infant face
    • infant left leg
    • infant face
    • infant right leg
    • infant face
    • infant left sholder
    • infant face
    • infant right sholder

bodyparts: MULTI
start: 0
stop: 1
numframes2pick: 2

Plotting configuration

skeleton_color: black
pcutoff: 0.05
dotsize: 5
alphavalue: 0.7
colormap: jet

Training,Evaluation and Analysis configuration


  • 0.95
    iteration: 2
    default_net_type: resnet_101
    default_augmenter: default
    snapshotindex: -1
    batch_size: 8

Cropping Parameters (for analysis and outlier frame detection)

cropping: false
croppedtraining: true
#if cropping is true for analysis, then set the values here:
x1: 0
x2: 640
y1: 277
y2: 624

Refinement configuration (parameters from annotation dataset configuration also relevant in this stage)


  • 50
  • 50

What can be the cause for this error? How can I fix it?

Another thing I’ve noticed is that in the training set, all points are duplicated. See attached image as an example.