You're currently viewing an old version of this dataset. To see the current version, click here.
EffiRLD_lodg01_summary.png
EffiRLD_lodg01_summary.png
- 名稱：模型摘要圖（Model summary）
- 格式：PNG
EffiRLD 模型摘要圖
一張圖看懂本模型，由上而下包含：
- 模型名稱、版本與授權
- 輸入影像與預測結果範例（獨立測試集圖磚 test_00663，2019 年臺中霧峰），附五類圖例
- 網路架構流程：EDANet 主幹、三個任務分支，以及 RBMTL 規則約束融合
- 獨立測試集效能（像素準確率、mIoU、倒伏稻 F1-score）、參數量，以及 GPU 與邊緣裝置推論速度
- 主幹網路出處與授權
圖中分割結果的顏色與模型輸出一致：黑色為背景、綠色為正常稻、灰色為道路、棕色為田埂、紅色為倒伏稻。
推論速度皆為單張、僅模型前向運算的實測值：GPU 為 NVIDIA RTX 5000 Ada、512×512、PyTorch FP32；邊緣裝置為 NVIDIA Jetson T5000、1536×1536、TensorRT FP16。
尺寸：768 × 1664 px。
EffiRLD model summary
The model at a glance, from top to bottom:
- Model name, version and license
- An example input image and its prediction (independent test-set tile test_00663, Wufeng, Taichung, 2019) with a five-class legend
- The network architecture: EDANet backbone, three task branches and the RBMTL rule-based fusion
- Independent test-set performance (pixel accuracy, mIoU, lodging F1-score), parameter count, and inference speed on a GPU and an edge device
- Backbone reference and license
The segmentation colors match the model output: black is background, green is normal rice, gray is road, brown is ridge and red is lodged rice.
Inference speeds are measured for a single image, model forward pass only: on the GPU, an NVIDIA RTX 5000 Ada with 512×512 input in PyTorch FP32; on the edge device, an NVIDIA Jetson T5000 with 1536×1536 input in TensorRT FP16.
Size: 768 × 1664 px.
Additional Information
| Field | Value |
|---|---|
| Data last updated | October 22, 2026 |
| Metadata last updated | October 22, 2026 |
| Created | October 22, 2026 |
| Format | PNG |
| Has views | True |
| Id | 3cf5123d-0b26-4069-b8d3-7d903f242ebc |
| Mimetype | image/png |
| Nchu ddl resource filetype | 基礎工具 |
| Package id | 84823116-08a7-4444-ac92-3152631710a1 |
| Position | 4 |
| Size | 494 KiB |
| State | active |