Nodes/comfyui_cv/CV EAST Text Detect
ComfyUI Node

CV EAST Text Detect

Detects scene text with the EAST CNN (cv2.dnn). Place the .onnx model in ComfyUI/models/onnx (e.g. east_text_detection_2026jul.onnx). Outputs axis-aligned bboxes (for the core 'Draw BBoxes' node), rotated quads (Nx4x2 corner points for 'CV Draw Polygon'), and scores. Zero detections is a valid result (empty outputs).

By bmad4ever·Created 3 months ago·Updated 2 days ago· 0
CV EAST Text Detect
  • image
  • bboxes
  • quads
  • scores
  • det_count
model
conf_threshold0.50
nms_threshold0.40
input_size320
Categoryimage/CV/dnn

Inputs (5)

NameTypeDefaultDescription
imageNPARRAY,IMAGEInput image. An IMAGE batch is processed frame by frame; an NPARRAY (BGR/RGB/gray, any dtype) is treated as one frame. Accepts a ComfyUI IMAGE/MASK directly (frame 0 of a batch) or an NPARRAY. Arithmetic ops (add, multiply, etc.) process the full IMAGE batch when both inputs have the same batch size.
modelCOMBOEAST .onnx model from ComfyUI/models/onnx.
conf_thresholdFLOAT0.500–1Minimum confidence to keep a detection. Lower detects more (and more false positives).
nms_thresholdFLOAT0.400–1Non-maximum suppression IoU threshold for merging overlapping text boxes.
input_sizeINT320320–1024Square input size for the network. Larger is more accurate but slower. 320 matches the reference demo.

Outputs (4)

NameTypeDescription
bboxesBOUNDING_BOXAxis-aligned {x, y, width, height, score, label} dicts - feed the core 'Draw BBoxes' node for a quick preview.
quadsNPARRAY(N, 4, 2) float32 rotated corner points per detection. Feed into 'CV Draw Polygon' for precise rotated outlines (set thickness=0 for filled polygons or a mask).
scoresNPARRAY(N,) float32 confidence per detection, same order as bboxes. Empty (0,) when nothing is detected.
det_countINTTotal number of text regions detected.