Datasets:

HoangTrungNguyen
/

Varroa_ASRM_training

Modalities:

Formats:

json

Size:

Libraries:

Dataset card Data Studio Files Files and versions

xet

Community

Dataset Viewer

Auto-converted to Parquet Duplicate

Split (1)

train · 1 rows

case string	seed int64	pretrained string	pretrained_transfer dict	epochs int64	image_size int64	batch_size int64	parameters int64	trainable_parameters int64	gflops float64	runtime_seconds float64	best_checkpoint string	learned_guidance dict	metrics dict	torch string	cuda string	gpu string
aduy_gap_factorized_k15_rgb_saturation_srm3_cls_edge_refine	43	yolov8n.pt	{ "total": 391, "direct": 229, "backbone_remapped": 162, "input_conv_adapted": 1, "method": "aduy_smart_transfer_plus_shifted_backbone_remap" }	200	640	8	1,603,284	0	0	551.095565	/marimo/three_dataset_fpn_p2_fusion_runtime/varroa/runs/paper_guidance_ablations_2seed_nomosaic/aduy_gap_factorized_k15_rgb_saturation_srm3_cls_edge_refine/seed_43/weights/best.pt	{ "model.22.srm.gamma_srm": -0.248291015625 }	{ "val": { "bbox_mAP": 0.26570095424585694, "bbox_mAP_50": 0.8009642689255505, "bbox_mAP_75": 0.08049195883691103, "precision": 0.8438087701991097, "recall": 0.8021043613718667 }, "test": { "bbox_mAP": 0.25766879932328296, "bbox_mAP_50": 0.8045606717159268, "bbox_mAP_75": 0.0677069...	2.11.0+cu130	13.0	NVIDIA RTX PRO 6000 Blackwell Server Edition

YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

Varroa: three YOLOv8-P2 ASRM variants

Standalone package for training the three current ASRM architectures on the Varroa dataset. It is isolated from LEVIR-Ship and from the original Varroa project.

Cases:

p2_asrm_fusion
input_asrm_auxiliary
input_asrm_guided_p2

The notebook follows Varroa/YOLO_custom/HF_YOLO_version_workflow.ipynb: gt_one, infected images only, class token 3 mapped to class 1, fixed conversion split seed 42, pretrained yolov8n.pt, image size 640, batch 16, 100 epochs, patience 20, workers 4 and training seeds 42/43.

The raw archive retains the original train/val/test layout. prepare_dataset.py deduplicates the positive CSV records and creates a reproducible 70/15/15 Ultralytics layout. All cases and training seeds use that same converted dataset.

Outputs:

test_metrics.json per case/seed;
comparison_varroa_asrm.csv with raw runs;
comparison_varroa_asrm_aggregate.csv with mean/std over seeds.

Version 2 exposes an auxiliary attention map but has no auxiliary loss, so its detection path is intentionally equivalent to the P2 baseline during this experiment.

Downloads last month: 43

Number of rows:

Total file size:

1.25 GB