Please provide the following information when requesting support.
• Hardware (Desktop/rtx 2070)
• Network Type (FPEnet)
• TLT Version (Please run “tlt info --verbose” and share “docker_tag” here)
• Training spec file(
sets: [train, validate]
gt_path: ‘annotations’
save_path: ‘tfrecords2’
gt_root_path: ‘/workspace/416x416_pictures/data/’
save_root_path: ‘/workspace/416x416_pictures/’
image_root_path: ‘/workspace/416x416_pictures/data/’
tfrecord_folder: ‘FpeTfRecords’
tfrecord_name: ‘data.tfrecords’
num_keypoints: 1
bbox_enlarge_ratio: 1.0
)
• How to reproduce the issue ?( on the docker I mentioned I run “fpenet dataset_convert -e /workspace/416x416_pictures/dataset_converter.yaml -r /workspace/416x416_pictures/tfrecords” and the resulting log:
2024-12-08 06:17:35,150 [TAO Toolkit] [INFO] nvidia_tao_tf1.cv.common.logging.logging 197: Log file already exists at /workspace/416x416_pictures/tfrecords/status.json
2024-12-08 06:17:35,151 [TAO Toolkit] [INFO] root 2102: Starting dataset convert.
/usr/local/lib/python3.8/dist-packages/nvidia_tao_tf1/cv/fpenet/scripts/dataset_convert.py:99: YAMLLoadWarning: calling yaml.load() without Loader=… is deprecated, as the default Loader is unsafe. Please read https://msg.pyyaml.org/load for full details.
args = load(f)
string indices must be integers
string indices must be integers
string indices must be integers
Json /workspace/416x416_pictures/data/annotations/train/annotatons.json has image count: 0
recordtype:/workspace/416x416_pictures/tfrecords2/train/FpeTfRecords/data.tfrecords count: 0
Set train has total image count: 0
DataIO for train done in 0.01 sec.
string indices must be integers
string indices must be integers
string indices must be integers
Json /workspace/416x416_pictures/data/annotations/validate/annotations.json has image count: 0
recordtype:/workspace/416x416_pictures/tfrecords2/validate/FpeTfRecords/data.tfrecords count: 0
Set validate has total image count: 0
DataIO for validate done in 0.0 sec.
2024-12-08 06:17:35,169 [TAO Toolkit] [INFO] root 2102: Dataset convert finished successfully.
Execution status: PASS
“annotations.json has image count: 0”
My annotatoio.json is so:
{
“image”: [
{
“id”: 0,
“filename”: “0.png”,
“width”: 80,
“height”: 80
},
{
“id”: 1,
“filename”: “1.png”,
“width”: 80,
“height”: 80
},
),],
“annotations”: [
{
“id”: 0,
“image_id”: 0,
“category_id”: 1,
“num_keypoints”: 1,
“keypoints”: [
46.66666666666667,
29.23076923076923,
2
],
“iscrowd”: 0,
“bbox”: [
0,
0,
80,
80
]
},
{
“id”: 1,
“image_id”: 1,
“category_id”: 1,
“num_keypoints”: 1,
“keypoints”: [
40.434782608695656,
34.94736842105263,
2
],
“iscrowd”: 0,
“bbox”: [
0,
0,
80,
80
]
}
],
“categories”: [
{
“id”: 1,
“name”: “rfx”,
“supercategory”: “none”,
“keypoints”: [
“midpoint”
],
“skeleton”:
}
]
}
I have 80x80 images where bbox is dummy “bbox”: [
0,
0,
80,
80
]… How should the annotations.json be so that it takes the images…? my folder structure:
/workspace//416x416_pictures/data
L__train
L__validate
L__annotations
__L__train
____L___annotations.json
__L__validate
____L___annotations.json
1 post - 1 participant