SwinTextSpotter
SwinTextSpotter copied to clipboard
mxin262
Pytorch re-implementation of Paper: SwinTextSpotter: Scene Text Spotting via Better Synergy between Text Detection and Text Recognition (CVPR 2022)
你好，我想问一下，我使用rects数据集来测试demo.py的结果，发现输出的output里没有汉字，我测试totaltext上的英文数据集就有显示英文的 以下是我输入的指令：使用的model为作者您提供的rects的pth文件 python demo/demo.py --config-file projects/SWINTS/configs/SWINTS-swin-chn_finetune.yaml --input input1.jpg --output ./output --confidence-threshold 0.4 --opts MODEL.WEIGHTS ./output/rects_model_final.pth SWINTS-swin-chn_finetune.yaml的配置： _BASE_: "Base-SWINTS_swin.yaml" MODEL: #WEIGHTS: "detectron2://ImageNetPretrained/torchvision/R-50.pkl" WEIGHTS: "./output/rects_model_final.pth" SWINTS: NUM_PROPOSALS: 300 NUM_CLASSES: 2 REC_HEAD: POOLER_RESOLUTION:...
训练的时候出现错误“RuntimeError: CUDA out of memory. Tried to allocate 20.00 MiB (GPU 0; 7.92 GiB total capacity; 6.76 GiB already allocated; 66.50 MiB free; 6.87 GiB reserved in total by PyTorch)...
Hi, I am trying to train LSVT dataset with this repo. Seems, you've been trained LSVT, https://github.com/mxin262/SwinTextSpotter/blob/e238a4b5d0c127480a838c6245c1e5e9eb2f9d59/detectron2/data/datasets/builtin.py#L58 but there is no transcription for that. is LSVT data trainable with this...
Traceback (most recent call last): File "demo/demo.py", line 99, in predictions, visualized_output = demo.run_on_image(img, args.confidence_threshold, path) File "C:\Users\Administrator\SwinTextSpotter\demo\predictor.py", line 67, in run_on_image vis_output = visualizer.draw_instance_predictions(predictions=instances, path=path) File "c:\users\administrator\swintextspotter\detectron2\utils\visualizer.py", line 413,...
I tried to train the model on VinText dataset and got this traceback after several iterations: ``` Traceback (most recent call last): File "/home/ccbien/projects/SceneText/exp/SwinTextSpotter/detectron2/engine/train_loop.py", line 140, in train self.run_step() File...
hi， I have another question，why one picture needs 6s to predict？ Comparing PAN++ 85FPS，What is our speed? I didnt find it in paper. Please tell me, thanks!
if I use the ctw1500 dataset, how to get the annotation file " instances_train2017.json" and "test_ctw1500_maxlen100.json", thanks.
