Detective-SAM-small-MVSS-Set

Official repository for Detective-SAM-small-MVSS-Set checkpoints, evaluated on AI-generated image detection and segmentation tasks.

📌 Available Versions

Version	Best Metric Score	Best Epoch	Updated
v1	0.2617562416485535	9	2026-11-06

• Active Version: v1
• Best Validation Score: 0.2617562416485535 (Epoch 9)
• Total Training Epochs: 9

---

📊 Master Cross-Evaluation Results

SID-UNet Master Cross-Evaluation Report

Overview

• Evaluated Checkpoints: 1
• Evaluated Configurations: 1
• Total Cross-Evaluation Runs: 1

Cross-Evaluation Matrices

Mean IoU (Intersection over Union) Cross-Evaluation Matrix

Checkpoint \ Config	eval_mvss_v1_protocol
detective_sam	0.0837

Dice / Pixel F1 Score Cross-Evaluation Matrix

Checkpoint \ Config	eval_mvss_v1_protocol
detective_sam	0.1122

AUROC (Pixel-level ROC-AUC) Cross-Evaluation Matrix

Checkpoint \ Config	eval_mvss_v1_protocol
detective_sam	0.5882

Pixel Accuracy Cross-Evaluation Matrix

Checkpoint \ Config	eval_mvss_v1_protocol
detective_sam	0.8568

Evaluation Total Loss Cross-Evaluation Matrix

Checkpoint \ Config	eval_mvss_v1_protocol
detective_sam	0.8923

Master Consolidated Ranking

Checkpoint	Config (Dataset)	Split	Samples	Loss	IoU	Dice / F1	AUROC	Pixel Acc	Precision	Recall	Specificity	Class Acc
detective_sam	eval_mvss_v1_protocol (KhangTruong/Tampered-Images-Combined-Dataset-Fake)	cross_test	17929	0.8923	0.0837	0.1122	0.5882	0.8568	0.2457	0.1031	0.9713	N/A

🖼️ Cross-Evaluation Heatmaps & Visual Illustrations

Cross Eval Iou Heatmap

Cross Eval F1 Heatmap

Cross Eval Pixel Acc Heatmap

Cross Eval Auroc Heatmap

Sample Predictions Mvss-V1-Protocol

---

Per-Checkpoint Deep-Dive

Checkpoint: detective_sam

• Path: outputs/RUN/default/checkpoints/checkpoint_best.pt

Config	Dataset	Loss	IoU	Dice / F1	AUROC	Pixel Acc	Samples
eval_mvss_v1_protocol	KhangTruong/Tampered-Images-Combined-Dataset-Fake	0.8923	0.0837	0.1122	0.5882	0.8568	17929

---

Per-Configuration / Dataset Deep-Dive

Configuration: eval_mvss_v1_protocol (Dataset: KhangTruong/Tampered-Images-Combined-Dataset-Fake)

• Config Path: configs/cross-eval/MVSS-v1-protocol.yaml

Checkpoint	Loss	IoU	Dice / F1	AUROC	Pixel Acc	Samples
detective_sam	0.8923	0.0837	0.1122	0.5882	0.8568	17929

Per-Dataset Breakdown: detective_sam on eval_mvss_v1_protocol

Sub-Dataset	IoU	Dice / F1	AUROC	Pixel Acc	Precision	Recall	Specificity	Samples
CASIA_V1	0.2234	0.2710	0.6183	0.9060	0.3544	0.2687	0.9680	920
CASIA_V2	0.3008	0.3530	0.6560	0.8654	0.4795	0.3361	0.9759	1000
COCO_Inpaint	0.0168	0.0277	0.5040	0.9319	0.0860	0.0243	0.9837	2000
CocoGlide	0.2187	0.2886	0.6083	0.7903	0.6224	0.2314	0.9853	507
Columbia_Gray	0.1270	0.1717	0.5460	0.7438	0.4252	0.1288	0.9633	180
Coverage	0.1341	0.1696	0.5684	0.8755	0.2621	0.1689	0.9679	100
HiFi_IFDL	0.0479	0.0755	0.5067	0.6897	0.2437	0.0581	0.9552	2000
IMD2020	0.0261	0.0396	0.5186	0.9620	0.0493	0.0633	0.9738	2000
IMD2020_RealLife	0.1450	0.1876	0.5844	0.9024	0.2693	0.2002	0.9686	1000
NIST16	0.0570	0.0942	0.3811	0.1740	0.6318	0.0662	0.6960	220
OpenSDID_Flux.1	0.0521	0.0765	0.5195	0.8330	0.2415	0.0587	0.9803	2000
OpenSDID_SD1.5	0.0646	0.0929	0.5276	0.8313	0.2716	0.0756	0.9796	2000
OpenSDID_SDXL	0.0522	0.0775	0.5208	0.8264	0.2473	0.0623	0.9793	2000
SAGI-D	0.2146	0.2696	0.6110	0.8589	0.3959	0.2351	0.9870	16
TGIF2_flux1dev	0.1012	0.1012	1.0000	0.9755	0.1012	0.1012	0.9755	1986

---

🚀 Quickstart & Checkpoint Usage

1. Pull checkpoints using sid-pull

To download the latest checkpoint directly to your local workspace:

sid-pull --name KhangTruong/Detective-SAM-small-MVSS-Set

To pull a specific version:

sid-pull --name KhangTruong/Detective-SAM-small-MVSS-Set --version v1

2. Resume or Fine-tune with sid-train

Once downloaded, resume training seamlessly:

sid-train --config configs/experiments/diffusion_diff_minimized/default.yaml --override training.epochs=40

Or specify explicit checkpoint path:

sid-train --resume checkpoints/checkpoint_latest.pt

3. Push new checkpoints using sid-push

sid-push --name KhangTruong/Detective-SAM-small-MVSS-Set --version v2

---

📂 Repository Layout

KhangTruong/Detective-SAM-small-MVSS-Set/
├── README.md               # Model Card & Documentation
├── config.yaml             # Model & Training Configuration
├── manifest.json           # Catalog of all versions & metrics
├── checkpoints/            # Active / Latest Checkpoints
│   ├── checkpoint_best.pt
│   ├── checkpoint_latest.pt
│   └── checkpoint_periodic.pt
├── reports/                # Evaluation reports & plots
│   ├── training_curves.png
│   └── test_evaluation_report.json
├── cross_evaluation/       # Master evaluation matrices & illustrations
│   ├── master_cross_evaluation_report.md
│   ├── cross_eval_matrix.json
│   └── illustrations/
└── versions/               # Version Archive
    └── v1/
        ├── checkpoints/
        ├── reports/
        └── cross_evaluation/