ComfyUI Qwen2.5-VL Object Detection Node

This repository provides a custom ComfyUI node for running object detection with the Qwen 2.5 VL model. The node downloads the selected model on demand, runs a detection prompt and outputs bounding boxes that can be used with segmentation nodes such as SAM2.

Nodes

DownloadAndLoadQwenModel

Downloads a chosen Qwen 2.5-VL model into models/Qwen and returns the loaded model and processor. You can choose which device to load the model onto (e.g. cuda:1 if you have multiple GPUs) and the precision for the checkpoint (INT4, INT8, BF16, FP16 or FP32). FlashAttention is used for FP16/BF16 but FP32 falls back to PyTorch SDPA since FlashAttention does not support it.

QwenVLDetection

Runs a detection prompt on an input image using the loaded model. The node outputs the raw text response and the bounding box coordinates. You can specify which boxes to return using the bbox_selection parameter and filter them with score_threshold:

• all – return all boxes (default)
• Comma-separated indices such as 0, 1,2 or 0,2 – return only the selected boxes, sorted by detection confidence
• merge_boxes – when enabled, merge the selected boxes into a single bounding box
• score_threshold – discard boxes whose confidence score is below this value (default 0)
• If no boxes remain after filtering, the highest scoring box is returned so downstream nodes like SAM2 do not fail

The bounding boxes are converted to absolute pixel coordinates so that they can be directly fed into the SAM2 nodes.

Usage

1. Place this repository inside your ComfyUI/custom_nodes directory.
2. From the Download and Load Qwen2.5-VL Model node, select the model you want to use, choose the desired precision (INT4/INT8/BF16/FP16/FP32) and, if necessary, choose the device (such as cuda:1) where it should be loaded. The snapshot download will resume automatically if a previous attempt was interrupted.
3. Connect the output model to Qwen2.5-VL Object Detection, provide an image and the object you want to locate (e.g. cat). Use bbox_selection to choose which boxes to keep (e.g. 0,2) and enable merge_boxes if you want them merged. Adjust score_threshold to filter out low-confidence results. The node will automatically build the detection prompt and return the selected boxes.
4. Feed the resulting bounding boxes into the SAM2 workflow for further processing.