Open Issues Need Help
View All on GitHubAI Summary: This GitHub issue addresses a medium-security vulnerability where RTSP URLs containing username and password are created and subsequently logged or exposed in error messages. This flaw could lead to sensitive credential leakage. The proposed solution involves implementing a custom log filter to redact credentials from logs and sanitizing exception messages to prevent exposure.
Real-time wildlife detection and telescope collision prevention system using RT-DETR + iNaturalist on NVIDIA A30
Real-time wildlife detection and telescope collision prevention system using RT-DETR + iNaturalist on NVIDIA A30
Real-time wildlife detection and telescope collision prevention system using RT-DETR + iNaturalist on NVIDIA A30
Real-time wildlife detection and telescope collision prevention system using RT-DETR + iNaturalist on NVIDIA A30
Real-time wildlife detection and telescope collision prevention system using RT-DETR + iNaturalist on NVIDIA A30
Real-time wildlife detection and telescope collision prevention system using RT-DETR + iNaturalist on NVIDIA A30
Real-time wildlife detection and telescope collision prevention system using RT-DETR + iNaturalist on NVIDIA A30
Real-time wildlife detection and telescope collision prevention system using RT-DETR + iNaturalist on NVIDIA A30
Real-time wildlife detection and telescope collision prevention system using RT-DETR + iNaturalist on NVIDIA A30
Real-time wildlife detection and telescope collision prevention system using RT-DETR + iNaturalist on NVIDIA A30
Real-time wildlife detection and telescope collision prevention system using RT-DETR + iNaturalist on NVIDIA A30
Real-time wildlife detection and telescope collision prevention system using RT-DETR + iNaturalist on NVIDIA A30
Real-time wildlife detection and telescope collision prevention system using RT-DETR + iNaturalist on NVIDIA A30
Real-time wildlife detection and telescope collision prevention system using RT-DETR + iNaturalist on NVIDIA A30
Real-time wildlife detection and telescope collision prevention system using RT-DETR + iNaturalist on NVIDIA A30
Real-time wildlife detection and telescope collision prevention system using RT-DETR + iNaturalist on NVIDIA A30
Real-time wildlife detection and telescope collision prevention system using RT-DETR + iNaturalist on NVIDIA A30
Real-time wildlife detection and telescope collision prevention system using RT-DETR + iNaturalist on NVIDIA A30
Real-time wildlife detection and telescope collision prevention system using RT-DETR + iNaturalist on NVIDIA A30
Real-time wildlife detection and telescope collision prevention system using RT-DETR + iNaturalist on NVIDIA A30
Real-time wildlife detection and telescope collision prevention system using RT-DETR + iNaturalist on NVIDIA A30
Real-time wildlife detection and telescope collision prevention system using RT-DETR + iNaturalist on NVIDIA A30
Real-time wildlife detection and telescope collision prevention system using RT-DETR + iNaturalist on NVIDIA A30
Real-time wildlife detection and telescope collision prevention system using RT-DETR + iNaturalist on NVIDIA A30
Real-time wildlife detection and telescope collision prevention system using RT-DETR + iNaturalist on NVIDIA A30
Real-time wildlife detection and telescope collision prevention system using RT-DETR + iNaturalist on NVIDIA A30
Real-time wildlife detection and telescope collision prevention system using RT-DETR + iNaturalist on NVIDIA A30
AI Summary: The current YOLOX inference system processes each camera's frames independently in separate threads, leading to inefficient GPU utilization due to small batch sizes (1). This issue proposes refactoring the architecture to batch frames from multiple cameras before sending them to the GPU, significantly reducing inference time and increasing GPU throughput. The goal is to transform 120 individual GPU calls per second into 30 batched calls, cutting GPU processing time by 66%.
Real-time wildlife detection and telescope collision prevention system using RT-DETR + iNaturalist on NVIDIA A30
Real-time wildlife detection and telescope collision prevention system using RT-DETR + iNaturalist on NVIDIA A30
AI Summary: The GitHub issue describes a performance problem where WebSocket detection messages are unnecessarily serialized and sent to clients, even when the underlying detection data hasn't changed meaningfully (e.g., empty frames, identical counts, minor bounding box movements). This leads to significant CPU overhead and network traffic, especially with multiple clients and high frame rates. The proposed solution involves implementing a change detection mechanism to only send messages when the detection data has truly changed.
Real-time wildlife detection and telescope collision prevention system using RT-DETR + iNaturalist on NVIDIA A30
Real-time wildlife detection and telescope collision prevention system using RT-DETR + iNaturalist on NVIDIA A30
Real-time wildlife detection and telescope collision prevention system using RT-DETR + iNaturalist on NVIDIA A30
Real-time wildlife detection and telescope collision prevention system using RT-DETR + iNaturalist on NVIDIA A30
Real-time wildlife detection and telescope collision prevention system using RT-DETR + iNaturalist on NVIDIA A30