D-FINE

D-FINE copied to clipboard

Is it possible to export ONNX model with 320x320 resolution?

8

Trying to run the model in an Intel iGPU, I'm trying to reduce de model to 320x320 resolution for real-time processing, but 640x640 is hardcoded in export_onnx library and if...

brainstormi

mAP is almost 0

4

**Describe the bug** mAP is almost 0 when train a 5 cls dataset finetune from coco or obj2coco pretrian mode such as: CUDA_VISIBLE_DEVICES=0,1,2,3 torchrun --master_port=7777 --nproc_per_node=4 train.py -c configs/dfine/custom/objects365/dfine_hgnetv2_m_obj2custom-building-defeat.yml --use-amp...

gjtjx

Question: Reducing False Positives with Empty annotation Images in training process

5

Question: Reducing False Positives with Empty Images Dear friends and code authors, Thank you for sharing this codebase and ideas! I'm seeing a high rate of false positives with my...

xuechaoyan

#Question: How we can process empty ground true boxes?

4

Suppose i have process image without ground true boxes (annotation file is empty). How i can correctly calculate loss and initialize ground true values?

KozlovKY

Hello, Thanks for your work. I was reading the D-FINE paper, and it is mentioned that D-FINE-L was trained for 72 epochs, but in the code, I found this https://github.com/Peterande/D-FINE/blob/4a1f73a8bcfac736a88abde9596d87f116d780a7/configs/dfine/dfine_hgnetv2_l_coco.yml#L34-L44...

SebastianJanampa

Instance Segmentation

6

Hi! I was able to train and fine tune on custom datasets and the performance is far better than other SOTA models (lmk if anyone needs help, more than happy...

Kavindu404

Early Stopping/Patience Epochs help needed

1

Any idea how to automatically stop training if your validation metric (AP) doesn't improve for patience epochs, saving compute resources and preventing overfitting?

FarhAnonymous

D-FINE is now available in 🤗 Transformers

6

Hi there, The D-FINE model is officially integrated in the Hugging Face Transformers library 🤗 It enables easy inference as well as fine-tuning on custom data. ## Resources * Models...

NielsRogge

does this nano model works with 640*640 img size can i give 1280*736 image size

1

does this nano model works with 640*640 img size can i give 1280*736 image size . if yes where is the parameter in config where i can change the img...

manoj-samal

Has anyone been converted the model to INT8 successfully?

3

Hi all, I couldn't convert the model from .pth -> ONNX -> TRT INT8. Has anyone met the same situation? It's successful to convert the model to FP16 or FP8....

horngjason