File size: 2,077 Bytes
fa9bc19 5c187e6 fa9bc19 add378b 5dd7e8a 925e670 fa9bc19 941f2e8 bec5b55 9e9f78e 1b4bf2b 9e9f78e 1b4bf2b fa9bc19 941f2e8 1b4bf2b 941f2e8 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 |
---
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>
|