diff --git "a/index.html" "b/index.html" --- "a/index.html" +++ "b/index.html" @@ -418,7 +418,7 @@ RepoFamilyTask - SizesNano mAP50 (ONNX)Nano mAP50-95 (INT8) + SizesNano mAP50-95 (ONNX)Nano mAP50-95 (INT8) Last modified @@ -442,7 +442,7 @@
-

mAP@0.5 — ONNX FP32 vs TFLite INT8

+

mAP@0.5-0.95 — ONNX FP32 vs TFLite INT8

@@ -452,6 +452,19 @@
+
+

Accuracy vs On-target Throughput

+ COCO val2017 · pipelined FPS measured on each platform by the EdgeFirst Profiler +
+
+

+ Each point is a public validation session: color = platform, shape = model family + (● YOLOv5 · ■ YOLOv8 · ▲ YOLO11 · ◆ YOLO26). X axis is sustained pipelined throughput on a log + scale — the full image load → decode → preprocess → inference → postprocess pipeline, not bare + inference time. ⚠ flagged sessions are shown at their measured accuracy. +

+
+

Nano Segmentation — Mask mAP

YOLOv8 / YOLO11 / YOLO26, COCO val2017 @@ -504,7 +517,8 @@ Det mAP50-95 Mask mAP50-95 Inf (ms) - Total (ms) + E2E (ms) + FPS Studio @@ -552,12 +566,37 @@
+
+

Platform Throughput Leaderboard

+ Best sustained pipelined throughput measured per platform · full image load → decode → preprocess → inference → postprocess pipeline +
+ + + + + + + + + +
#PlatformFastest modelFPS (median)E2E latency (ms)
+

+ Throughput is the measured pipelined rate — stages overlap across frames, so it exceeds + 1000 ÷ end-to-end latency. Measured by the + EdgeFirst Profiler. +

+
+

Per-platform Leaderboards

- - + + +
+
+ +
@@ -583,7 +622,7 @@
- +