search

Irene323 / GFTE

A GCN-based table structure recognition method
222 stars 54 forks source link

How to run prediction on new PDF document #4

Open Sharathmk99 opened 4 years ago

Sharathmk99 commented 4 years ago

Hi,

Thank you for your amazing paper and source code. I was able to successfully run your code. Based on my understanding, input will be table cell bounding box, table position(row, col) and y will be its row or cell. Is my understanding correct?

If i have new PDF document which contains only Table(Assume i have cropped table from document). How can use GFTE to get rows and col of the table?

Thank you!!

cccusername commented 4 years ago

@Sharathmk99 hello, I run this code but i meet some trouble as follow: TypeError: knn_graph() takes from 2 to 6 positional arguments but 7 were given could you tell me which vision of pytorch and torch_geometric you use, thank you!

jingmouren commented 4 years ago

https://github.com/rusty1s/pytorch_geometric/issues/1521 doesn't solve. @Sharathmk99. what's your torch_geometric and torch_cluster version?

BrowningWan commented 4 years ago

@cccusername Just delete the arguments "num_workers" in line 129 from torch_geometric.nn.pool.init.py.

Sharathmk99 commented 4 years ago

@jingmouren Both are latest versions

harirajeev commented 4 years ago

@Sharathmk99 , where you able to run inference on the code , for a new pdf document ?

Sharathmk99 commented 4 years ago

@harirajeev I didn't try again. Because i think we also need one more prediction if two words belongs to same cells.

I started with TIES-2.0 repository, but its based on tensorflow. I'll keep you updated if i was able to solve.

Sharathmk99 commented 3 years ago

In meantime if anyone solves the problem, please update here.