| ## BUDDI Table Factory: A toolbox for generating synthetic documents with annotated tables and cells |
|
|
|
|
| **About** |
|
|
| In table detection, we initialize the weights with a pre-trained CDeCNet model using COCO dataset. We re-train the model for five epochs using a stochastic gradient descent optimizer with a learning rate of 0.00125, the momentum of 0.9, and weight decay of 0.0001. |
|
|
| ***Hardware Used*** |
|
|
| We perform all the experiments on NVIDIA GeForce RTX 2080 Ti GPU with 12 GB GPU memory, Intel(R) Xeon(R) CPU E5-2640 v2 @ 2.00GHz, and 128 GB of RAM. |
|
|
| **Table Detection Model & Training Parameter** |
| ***Optimizer*** |
| | Parameter |Value | |
| |--|--| |
| | Type | SGD | |
| | Learning Rate |0.00125 | |
| | Momentum | 0.8 | |
| | Weight Decay |0.001 | |
|
|
|
|
| *** Learning Policy *** |
| | Parameter |Value | |
| |--|--| |
| | Policy | Step | |
| |Warmup | Linear | |
| | Warmup Iteration | 100 | |
| | Warmup Ratio |0.001 | |
| | Step | 4,16,32 | |
|
|
| ***General Parameter*** |
| | Parameter |Value | |
| |--|--| |
| | Epoch | 5 | |
| | Step Interval |50 | |
| |
| ***Model Paper Reference*** |
| |
| CDeC-Net: Composite Deformable Cascade Network for Table Detection in Document Images |
|
|
| https://arxiv.org/abs/2008.10831 |
