Empyt classifications with AI4GAmazonRainforest

[MattB-SF]

Search before asking

• I have searched the Pytorch-Wildlife issues and found no similar bug report.

Bug

Hi there,

I would be interested in using AI4GAmazonRainforest Classification Model, but unforunately the resulting json files always has completely empty columns for the classification. The detection works though.
I am not sure what I am doing wrong. I have tried various different images from our Camera Trap project in Bolivia, but sofar without any success.
My code looks likes this:

def batch_detect_classify(image_folder, md_output_file, class_output_file, model_file):
from PytorchWildlife.models import detection as pw_detection
from PytorchWildlife.models import classification as pw_classification
from PytorchWildlife.data import datasets as pw_data
from PytorchWildlife.data import transforms as pw_trans
from PytorchWildlife import utils as pw_utils
import torch
from torch.utils.data import DataLoader
from PIL import ImageFile
ImageFile.LOAD_TRUNCATED_IMAGES = True

DEVICE = "cuda" if torch.cuda.is_available() else "cpu"

#' Initializing the MegaDetectorV6 model for image detection
detection_model = pw_detection.MegaDetectorV6(device=DEVICE, pretrained=True, version=model_file)

#' Classification model
classification_model = pw_classification.AI4GAmazonRainforest(device=DEVICE, version='v2')

#' Performing batch detection on the images
det_results = detection_model.batch_image_detection(image_folder, batch_size=16)

pw_utils.save_detection_json(det_results, md_output_file, categories=detection_model.CLASS_NAMES,
exclude_category_ids=[], # Category IDs can be found in the definition of each model.
exclude_file_path=None)

clf_results = classification_model.batch_image_classification(det_results=det_results, id_strip=image_folder)

#' Saving the detection results in timelapse JSON format
pw_utils.save_detection_classification_timelapse_json(det_results=det_results,
clf_results=clf_results,
det_categories=detection_model.CLASS_NAMES,
clf_categories=classification_model.CLASS_NAMES,
output_path=class_output_file)

Any ideas why the classification result turns out empty is highly appreciated.
Thanks,

Matt

Environment

No response

Minimal Reproducible Example

No response

Additional

No response

Are you willing to submit a PR?

• Yes I'd like to help by submitting a PR!