Unverified50% confidenceBenchmarkExact time
GTR-Depth achieves a single-frame inference latency of 2.7 milliseconds on the NVIDIA DRIVE AGX Thor platform.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/4/2026
First Seen
Valid until: 1/2/2027
Sources
Related Entities
Related Claims
UnverifiedIf VSR is run using only a CPU, the speed gap compared to a GPU can be as large as tens of times.64% similarVerifiedNVIDIA's TensorRT-LLM deeply optimizes inference speed for its hardware architecture.64% similarUnverifiedNVIDIA's Instant-NGP (2022) compressed NeRF training time from hours to seconds through multi-resolution hash encoding.63% similarUnverified使用 PaddleOCR、EasyOCR 等深度学习 OCR 引擎识别准确率提升到 95% 以上，但单帧处理时间增加到 1-2 秒且需要 GPU 加速63% similarUnverifiedNVIDIA NIM内置了TensorRT-LLM等推理加速引擎62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/971171API
curl https://kongchang.com/api/v1/knowledge/claims/971171MCP
get_claim(id=971171)