Inconsistency with extracted data VS annotation tool

[jutanke]

If we run the following code with the latest data annotation extraction

from PythonAPI.bam import BAM

dataset_dir = "/mnt/Data/datasets/bonn_activity_maps"
dataset_name = "190726"

bam = BAM(dataset_dir, dataset_name, image_format="png")

all_activ = bam.get_activities_for_person(pid=1)

all_walking_activ = list(
    sorted(
        [act for act in all_activ if act["label"] == "walking"],
        key=lambda ac: ac["start_frame"],
    )
)

print("first walking of actor 1", len(all_walking_activ), all_walking_activ[0])

we get: {'label': 'walking', 'start_frame': 45, 'end_frame': 109}

However, this does not coincide with what the annotation tool says:

Even more curiously, add frame 45 there is no person annotated at all:

[jutanke]

This is due to downsampling