Abstract
A novel framework named Markov Clustering Network
(MCN)is proposed for fast and robust scene text detec-
tion.MCN predicts instance-level bounding boxes by firstly
converting an image into a Stochastic Flow Graph(SFG)
and then performing Markov Clustering on this graph.Our
method can detect text objects with arbitrary size and orien-
tation without prior knowledge of object size.The stochas-
tic flow graph encode objects'local correlation and se-
mantic information.An object is modeled as strongly con-
nected nodes,which allows flexible bottom-up detection for
scale-varying and rotated objects.MCN generates bound-
ing boxes without using Non-Maximum Suppression,and it
can be fully parallelized on GPUs.The evaluation on public
benchmarks shows that our method outperforms the existing
methods by a large margin in detecring multioriented text
objects.MCN achieves new state-of-art performance on
challenging MSRA-TD500 dataset with precision of 0.88,
recall of O.79 and F-score of 0.83.Also,MCN achieves re-
altime inference with frame rate of 34 FPS,which is 1.5x
speedup when compared with the fastest scene text detection
algorithm.