Predict and Analyse YOLOv8 model on xView dataset.

[bshakya77]

Hello,
I am trying to predict and analyze the YOLOv8n model on xview dataset. I am using the baseline YOLOv8n model pretrained on COCO dataset. I used the following command:
!sahi predict \ --source ./datasets/xView/sliced/subsets/subsets_val_640_50 \ --dataset_json_path ./datasets/xView/sliced/subsets/annotations/subsets_val_640_50.json \ --model_type yolov8 \ --model_path ./models/baseline/yolov8n.pt \ --slice_height 256 \ --slice_width 256 \ --overlap_height_ratio 0.5 \ --overlap_width_ratio 0.5

It does not predict anything and mAP is all 0.

However when I used the YOLOv8n model trained on xview dataset. It predicts the objects and mAP is as follows on subset of 50 images:

My question: is it necessary to train the model on the target dataset before prediction and evaluation or we can use the COCO trained model using transfer learning.

Thank you.

Regards,
Bijay