why this code caltech is consider only one label? about dataset_to_voc_converter HOT 5 OPEN

@zeynali Just modify this line of code to support person and people detection

from dataset_to_voc_converter.

@zeynali I updated my code and just specify the person_types argument for both 'person' and 'people' data.

from dataset_to_voc_converter.

@CasiaFan ,Thanks , but what's different between ["person", "person-fa", "person?"] classes ?
this code incorrect convert to xml file : after converting finished , please see the xml file.

annotations/set00/bbox/set00_V013_1511.xml

in this xml file , this converted that have 6 people , In the event that , if you see associated with this JPG , you see this have 3 person and 2 people.
problem's is that this can't convert xml with two different annotations , that's xmls file are or Person or People , Even though there are several person or people in a JPG, they all stick to one of them.

from dataset_to_voc_converter.

@zeynali In VOC annotation format, all categories in one image will appear in single xml annotation file. If you want to separate each category into different annotation files just in one run, just fork and modify the script freely.
As for the caltech person types, person means individual, person-fa means person in distance, person? means cropped region may be a person.

from dataset_to_voc_converter.

@CasiaFan , i want to know , in the original caltech dataset annotations (seq files) , Originally those team annotations each images in the one class or when we convert these seq file to xml Makes it all objects in each image be one class ?

from dataset_to_voc_converter.