Instructions to use cmes-deepvision/ACR-instance-segmentation-RF-FP16-v1.0.0 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- TensorRT
How to use cmes-deepvision/ACR-instance-segmentation-RF-FP16-v1.0.0 with TensorRT:
# No code snippets available yet for this library. # To use this model, check the repository files and the library's documentation. # Want to help? PRs adding snippets are welcome at: # https://github.com/huggingface/huggingface.js
- Notebooks
- Google Colab
- Kaggle
ACR-instance-segmentation-RF(FP16)-v1.0.0
ACR instance segmentationμ© cmes_RF_960_v1.0.0 RF-DETR Seg2XLarge λͺ¨λΈμ
RTX 5070κ³Ό RTX 5080μμ κ°κ° λΉλν TensorRT FP16 λ°°ν¬ ν¨ν€μ§μ
λλ€.
Hugging Face μ μ₯μ IDλ κ΄νΈλ₯Ό νμ©νμ§ μμΌλ―λ‘ μ μ₯μ μ΄λ¦μ
cmes-deepvision/ACR-instance-segmentation-RF-FP16-v1.0.0μ μ¬μ©ν©λλ€.
μμ§ μ ν
| λμ GPU | μμ§ νμΌ | ν¬κΈ° | SHA256 |
|---|---|---|---|
| RTX 5070 | engines/cmes_RF_960_v1.0.0.fp16.RTX5070.trt |
81,984,724 bytes | 456d4855706691f434eed48ab8950640f6be200ab94d8b74d62183f269911c92 |
| RTX 5080 | engines/cmes_RF_960_v1.0.0.fp16.RTX5080.trt |
82,088,348 bytes | 667d2bc084092501f5e5bd1391cff76aaa640b24e28b2aaa18effd536498c577 |
TensorRT serialized engineμ GPUμ TensorRT λ²μ μ μ’ μλ©λλ€. RTX 5070μμλ RTX 5070 μμ§μ, RTX 5080μμλ RTX 5080 μμ§μ μ¬μ©νμμμ€. λ€λ₯Έ GPUλ‘ μμ§μ 볡μ¬ν΄ μ¬μ©νλ κ²μ μ§μνμ§ μμ΅λλ€.
κ²μ¦ νκ²½
- NVIDIA driver:
580.173.02 - CUDA runtime used by PyTorch:
12.8 - PyTorch:
2.11.0+cu128 - TensorRT:
10.16.1.11 - compute capability:
12.0 - μ
λ ₯: RGB, batch 1,
1 x 3 x 960 x 960 - κΈ°λ³Έ confidence threshold:
0.50
μμ§ λ΄λΆ κ³μ°μ FP16μ΄κ³ μ μΆλ ₯ bindingμ FP32μ λλ€.
μ€μΉ
python -m pip install -r requirements.txt
λ€μ΄λ‘λ
from huggingface_hub import hf_hub_download
engine_path = hf_hub_download(
repo_id="cmes-deepvision/ACR-instance-segmentation-RF-FP16-v1.0.0",
filename="engines/cmes_RF_960_v1.0.0.fp16.RTX5080.trt", # GPUμ λ§κ² λ³κ²½
)
μΆλ‘
μ μ₯μ μ 체λ₯Ό λ°μ λ€ λ€μκ³Ό κ°μ΄ μ€νν©λλ€. --engineμ μλ΅νλ©΄ νμ¬ GPU
μ΄λ¦μμ RTX 5070/5080 μμ§μ μλμΌλ‘ μ νν©λλ€.
python scripts/infer_tensorrt.py image.jpg --threshold 0.50
μμ§μ λͺ μνλ €λ©΄:
python scripts/infer_tensorrt.py image.jpg \
--engine engines/cmes_RF_960_v1.0.0.fp16.RTX5070.trt \
--output-dir outputs
μΆλ ₯ ν΄λμλ mask/bbox overlay JPEGμ detection JSONμ΄ μμ±λ©λλ€.
FP16 λ³ν μμ μ±
Test_v2.2.1μ 50κ° νμ€ν¬μμ κ³ μ λν μ΄λ―Έμ§ 1μ₯μ© μ΄ 50μ₯μ μ¬μ©νμ΅λλ€.
confidence 0.50, ν΄λμ€ μΌμΉ Hungarian matching, bbox/mask IoU >= 0.50
쑰건μμ PyTorch FP16 κ²°κ³Όμ TensorRT FP16 κ²°κ³Όλ₯Ό λΉκ΅νμ΅λλ€.
| GPU | BBox Precision | BBox Recall | Mask Precision | Mask Recall | νμ |
|---|---|---|---|---|---|
| RTX 5080 | 98.33% | 98.88% | 98.44% | 98.99% | PASS |
| RTX 5070 | 98.33% | 98.66% | 98.44% | 98.77% | PASS |
ν΅κ³Ό κΈ°μ€μ λ€ μ§νκ° λͺ¨λ 98% μ΄μμΈ κ²½μ°μ λλ€. μ΄λ² λν νλ³Έμμλ FP16 λ³νμΌλ‘ μΈν μ μλ―Έν μ νλ μ νκ° νμΈλμ§ μμμ΅λλ€.
μλ
warm-up 5μ₯ μ΄ν 50μ₯ x 3ν, batch 1λ‘ μΈ‘μ νμ΅λλ€. E2Eλ λ©λͺ¨λ¦¬ λ΄ PIL μ΄λ―Έμ§μ μ μ²λ¦¬, μΆλ‘ , segmentation νμ²λ¦¬μ CPU NumPy λ³νμ ν¬ν¨νλ©° λμ€ν¬ I/Oμ λͺ¨λΈ/μμ§ λ‘λλ μ μΈν©λλ€.
| GPU | PyTorch FP16 E2E | TensorRT FP16 E2E | TensorRT FPS | CT κ°μ | μμ§ λ¨λ |
|---|---|---|---|---|---|
| RTX 5080 | 28.48 ms | 16.87 ms | 59.28 | 40.77% | 6.63 ms / 150.72 FPS |
| RTX 5070 | 40.68 ms | 24.72 ms | 40.45 | 39.23% | 10.78 ms / 92.74 FPS |
μλ³Έ μΈ‘μ μΉλ benchmarks/μ ν¬ν¨λμ΄ μμ΅λλ€.
ν΄λμ€ μμ
0 dropping item
1 dumping item
2 item in pb bag
3 item in tote
4 item on buffer
5 item on floor
6 item on plate
7 item out of buffer
8 item out of tote
9 picked item
10 tote
νκ° λ²μ
μ΄ μ μ₯μμ GPUλ³ κ²μ¦μ λ³ν μμ μ±μ νμΈνκΈ° μν 50μ₯ λν νλ³Έμ
λ°±μλ μΌμΉλ νκ°μ
λλ€. λͺ¨λΈ μ체μ μ 체 6,475μ₯ GT κΈ°λ° μ±λ₯μ
ACR-instance-segmentation-RF-v1.0.0μ
νκ° κ²°κ³Όλ₯Ό μ°Έκ³ νμμμ€. λ³Έ μ μ₯μμ μΌμΉλ μμΉλ₯Ό μ 체 ν
μ€νΈμ
mAPλ‘
ν΄μνλ©΄ μ λ©λλ€.
- Downloads last month
- -