Add/update model card for small

README.md

@@ -21,12 +21,12 @@ This repo contains all four standard quantizations of this variant. **F16 is the
 
 | File | Quant | Size (MB) | Recall @ IoU 0.5 | Recall @ IoU 0.95 | Mean \|Δscore\| | Latency (median ms, T=8) |
 |---|---|---:|---:|---:|---:|---:|
-| `rfdetr-small-f32.gguf` | F32 | 119.0 | 1.0000 | 1.0000 | 0.0021 | 100.1 |
-| `rfdetr-small-f16.gguf` ← **recommended** | F16 | 64.0 | 1.0000 | 1.0000 | 0.0012 | 94.3 |
-| `rfdetr-small-q8_0.gguf` | Q8_0 | 38.2 | 1.0000 | 1.0000 | 0.0098 | 110.2 |
-| `rfdetr-small-q4_K.gguf` | Q4_K | 31.2 | 0.8333 | 0.5000 | 0.0164 | 116.0 |
+| `rfdetr-small-f32.gguf` | F32 | 119.0 | 0.9762 | 0.9762 | 0.0012 | 100.1 |
+| `rfdetr-small-f16.gguf` ← **recommended** | F16 | 64.0 | 0.9762 | 0.9762 | 0.0013 | 94.3 |
+| `rfdetr-small-q8_0.gguf` | Q8_0 | 38.2 | 0.9476 | 0.9476 | 0.0041 | 110.2 |
+| `rfdetr-small-q4_K.gguf` | Q4_K | 31.2 | 0.9238 | 0.8286 | 0.0206 | 116.0 |
 
-All accuracy numbers above are computed against the upstream PyTorch reference (`rfdetr 1.9.0`) on 1 image (test_image.jpg) at threshold 0.5. Latency is measured with `rfdetr-cli bench` (8 iters + 3 warmup) at T=8 threads on a single Intel Core i7-12800HX image (test_image.jpg).
+All accuracy numbers above are computed against the upstream PyTorch reference (`rfdetr 1.9.0`) on 7 images (000000000139.jpg, 000000000632.jpg, 000000039769.jpg, 000000087038.jpg, 000000252219.jpg, 000000397133.jpg, bus.jpg) at threshold 0.5. Latency is measured with `rfdetr-cli bench` (8 iters + 3 warmup) at T=8 threads on a single Intel Core i7-12800HX image (000000000139.jpg, 000000000632.jpg, 000000039769.jpg, 000000087038.jpg, 000000252219.jpg, 000000397133.jpg, bus.jpg).
 
 ## Architecture
 
@@ -71,7 +71,7 @@ build/bin/rfdetr-cli detect \
 
 ## Accuracy methodology
 
-All accuracy metrics are computed against the upstream PyTorch reference (`rfdetr 1.9.0`) on 1 image (test_image.jpg) at threshold 0.5. Each detection match uses greedy Hungarian-style assignment by IoU (≥ 0.5 lenient, ≥ 0.95 strict) with class equality required.
+All accuracy metrics are computed against the upstream PyTorch reference (`rfdetr 1.9.0`) on 7 images (000000000139.jpg, 000000000632.jpg, 000000039769.jpg, 000000087038.jpg, 000000252219.jpg, 000000397133.jpg, bus.jpg) at threshold 0.5. Each detection match uses greedy Hungarian-style assignment by IoU (≥ 0.5 lenient, ≥ 0.95 strict) with class equality required.
 
 See [BENCHMARK.md](https://github.com/adithyab94/rf-detr.cpp/blob/main/BENCHMARK.md) and [`benchmarks/results/accuracy_sweep.json`](https://github.com/adithyab94/rf-detr.cpp/blob/main/benchmarks/results/accuracy_sweep.json) for the full sweep across the (variant × quant) cells.
 
@@ -79,7 +79,7 @@ See [BENCHMARK.md](https://github.com/adithyab94/rf-detr.cpp/blob/main/BENCHMARK
 
 - Source project: [Roboflow RF-DETR](https://github.com/roboflow/rf-detr)
 - Upstream package: `rfdetr==1.9.0`
-- Converted with [rfdetr.cpp](https://github.com/adithyab94/rf-detr.cpp) at commit [`f0fd7f4ff643`](https://github.com/adithyab94/rf-detr.cpp/commit/f0fd7f4ff643824862071396cafeaf691f3cc800)
+- Converted with [rfdetr.cpp](https://github.com/adithyab94/rf-detr.cpp) at commit [`69d133e1e979`](https://github.com/adithyab94/rf-detr.cpp/commit/69d133e1e979c2dbf547507dfd8a5c5cec21e1f6)
 - Checkpoint: official pretrained `rfdetr-small` weights (downloaded by the `rfdetr` package on first use)
 
 ### Checksums (SHA-256)