|
--- |
|
tags: |
|
- image-segmentation |
|
library_name: keras |
|
--- |
|
## Model description |
|
The original idea from Keras examples [Monocular depth estimation](https://keras.io/examples/vision/depth_estimation/) of author [Victor Basu](https://www.linkedin.com/in/victor-basu-520958147/) |
|
|
|
Full credits go to [Vu Minh Chien](https://www.linkedin.com/in/vumichien/) |
|
|
|
Depth estimation is a crucial step towards inferring scene geometry from 2D images. The goal in monocular depth estimation is to predict the depth value of each pixel or infer depth information, given only a single RGB image as input. |
|
|
|
## Dataset |
|
[NYU Depth Dataset V2](https://cs.nyu.edu/~silberman/datasets/nyu_depth_v2.html) is comprised of video sequences from a variety of indoor scenes as recorded by both the RGB and Depth cameras from the Microsoft Kinect. |
|
|
|
## Training procedure |
|
|
|
### Training hyperparameters |
|
**Model architecture**: |
|
- UNet with a pretrained DenseNet 201 backbone. |
|
|
|
The following hyperparameters were used during training: |
|
- learning_rate: 1e-04 |
|
- train_batch_size: 16 |
|
- seed: 42 |
|
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 |
|
- lr_scheduler_type: ReduceLROnPlateau |
|
- num_epochs: 10 |
|
|
|
### Training results |
|
|
|
| Epoch | Training loss | Validation Loss | Learning rate | |
|
|:------:|:-------------:|:---------------:|:-------------:| |
|
| 1 | 0.1333 | 0.1315 | 1e-04 | |
|
| 2 | 0.0948 | 0.1232 | 1e-04 | |
|
| 3 | 0.0834 | 0.1220 | 1e-04 | |
|
| 4 | 0.0775 | 0.1213 | 1e-04 | |
|
| 5 | 0.0736 | 0.1196 | 1e-04 | |
|
| 6 | 0.0707 | 0.1205 | 1e-04 | |
|
| 7 | 0.0687 | 0.1190 | 1e-04 | |
|
| 8 | 0.0667 | 0.1177 | 1e-04 | |
|
| 9 | 0.0654 | 0.1177 | 1e-04 | |
|
| 10 | 0.0635 | 0.1182 | 9e-05 | |
|
|
|
|
|
|
|
### View Model Demo |
|
|
|
 |
|
|
|
|
|
<details> |
|
|
|
<summary> View Model Plot </summary> |
|
|
|
 |
|
|
|
</details> |
|
|
