Ultralytics
ultralytics ultralytics Q-a Discussions
Q&A Discussions
Ask the community for help
-
You must be logged in to vote Does the Ultralytics Python Package Transmit Any Data From Local Devices by Default?
documentationImprovements or additions to documentation questionFurther information is requested -
You must be logged in to vote YOLO World has extremely low training metrics on LVIS.
bugSomething isn't working as intended in the official Ultralytics package. detectObject Detection issues, PR's -
You must be logged in to vote Building off pre existing yolov8n-pose.pt
enhancementNew feature or request posePose/keypoint estimation models -
You must be logged in to vote Implementing MobileNetV2 backbone into YOLO26n
questionFurther information is requested detectObject Detection issues, PR's embeddedPertaining to IOT, low-power or edge devices such as Jetson, Rasberry Pi etc. -
You must be logged in to vote Does labelling the whole object help in better detection of the parts?
questionFurther information is requested detectObject Detection issues, PR's -
You must be logged in to vote YOLO misclassifies simulated objects (Gazebo) due to COCO domain gap. Best practical fix without full retraining?
enhancementNew feature or request detectObject Detection issues, PR's embeddedPertaining to IOT, low-power or edge devices such as Jetson, Rasberry Pi etc. -
You must be logged in to vote what dose pipeline do
devopsGitHub Devops or MLops -
You must be logged in to vote Training Yolo on a custom 3 class dataset
detectObject Detection issues, PR's -
You must be logged in to vote SAM3: Passing visual prompt from one image to another
segmentInstance Segmentation issues, PR's -
You must be logged in to vote /question about yolo model train
questionFurther information is requested -
You must be logged in to vote -
You must be logged in to vote -
You must be logged in to vote YOLO26 TFLite — Manual inference gives different results than Ultralytics' predict()
questionFurther information is requested detectObject Detection issues, PR's exportsModel exports (ONNX, TensorRT, TFLite, etc.) -
You must be logged in to vote -
You must be logged in to vote -
You must be logged in to vote How to train and deploy an ultra-small YOLO model on a constrained device (OpenArtPlus v1.0, 2MB SRAM, 64MB SDRAM)?
embeddedPertaining to IOT, low-power or edge devices such as Jetson, Rasberry Pi etc. exportsModel exports (ONNX, TensorRT, TFLite, etc.) -
You must be logged in to vote Power consumption measurement
questionFurther information is requested -
You must be logged in to vote -
You must be logged in to vote Clarification regarding mAP discrepancies between official benchmarks and local validation for YOLO11 and YOLO26
documentationImprovements or additions to documentation questionFurther information is requested detectObject Detection issues, PR's -
You must be logged in to vote Is it possible to enable AMP on ROCm/Windows?
questionFurther information is requested dependenciesDependencies and packages -
You must be logged in to vote Reducing small object false positives with high-resolution training / Hyperparameter Tuning
questionFurther information is requested detectObject Detection issues, PR's -
You must be logged in to vote I trained the model with 2026, but detected 2 same boxes on one object
questionFurther information is requested detectObject Detection issues, PR's
