For this project, we will be using data from the Waymo Open dataset.
[OPTIONAL] - The files can be downloaded directly from the website as tar files or from the Google Cloud Bucket as individual tf records. We have already provided the data required to finish this project in the workspace, so you don't need to download it separately.
The data you will use for training, validation and testing is organized as follow:
/home/workspace/data/waymo
- training_and_validation - contains 97 files to train and validate your models
- train: contain the train data (empty to start)
- val: contain the val data (empty to start)
- test - contains 3 files to test your model and create inference videos
The training_and_validation folder contains file that have been downsampled: we have selected one every 10 frames from 10 fps videos. The testing folder contains frames from the 10 fps video without downsampling.
You will split this `training_and_validation` data into `train`, and `val` sets by completing and executing the `create_splits.py` file.
### Experiments
The experiments folder will be organized as follow:
experiments/ - pretrained_model/ - exporter_main_v2.py - to create an inference model - model_main_tf2.py - to launch training - reference/ - reference training with the unchanged config file - experiment0/ - create a new folder for each experiment you run - experiment1/ - create a new folder for each experiment you run - experiment2/ - create a new folder for each experiment you run - label_map.pbtxt ...
## Prerequisites
### Local Setup
For local setup if you have your own Nvidia GPU, you can use the provided Dockerfile and requirements in the [build directory](./build).
Follow [the README therein](./build/README.md) to create a docker container and install all prerequisites.
### Download and process the data
**Note:** ”If you are using the classroom workspace, we have already completed the steps in the section for you. You can find the downloaded and processed files within the `/home/workspace/data/preprocessed_data/` directory. Check this out then proceed to the **Exploratory Data Analysis** part.
The first goal of this project is to download the data from the Waymo's Google Cloud bucket to your local machine. For this project, we only need a subset of the data provided (for example, we do not need to use the Lidar data). Therefore, we are going to download and trim immediately each file. In `download_process.py`, you can view the `create_tf_example` function, which will perform this processing. This function takes the components of a Waymo Tf record and saves them in the Tf Object Detection api format. An example of such function is described [here](https://tensorflow-object-detection-api-tutorial.readthedocs.io/en/latest/training.html#create-tensorflow-records). We are already providing the `label_map.pbtxt` file.
You can run the script using the following command:
python download_process.py --data_dir {processed_file_location} --size {number of files you want to download}
You are downloading 100 files (unless you changed the `size` parameter) so be patient! Once the script is done, you can look inside your `data_dir` folder to see if the files have been downloaded and processed correctly.
### Classroom Workspace
In the classroom workspace, every library and package should already be installed in your environment. You will NOT need to make use of `gcloud` to download the images.
## Instructions
### Exploratory Data Analysis
You should use the data already present in `/home/workspace/data/waymo` directory to explore the dataset! This is the most important task of any machine learning project. To do so, open the `Exploratory Data Analysis` notebook. In this notebook, your first task will be to implement a `display_instances` function to display images and annotations using `matplotlib`. This should be very similar to the function you created during the course. Once you are done, feel free to spend more time exploring the data and report your findings. Report anything relevant about the dataset in the writeup.
Keep in mind that you should refer to this analysis to create the different spits (training, testing and validation).
### Create the training - validation splits
In the class, we talked about cross-validation and the importance of creating meaningful training and validation splits. For this project, you will have to create your own training and validation sets using the files located in `/home/workspace/data/waymo`. The `split` function in the `create_splits.py` file does the following:
* create three subfolders: `/home/workspace/data/train/`, `/home/workspace/data/val/`, and `/home/workspace/data/test/`
* split the tf records files between these three folders by symbolically linking the files from `/home/workspace/data/waymo/` to `/home/workspace/data/train/`, `/home/workspace/data/val/`, and `/home/workspace/data/test/`
Use the following command to run the script once your function is implemented:
python create_splits.py --data-dir /home/workspace/data
### Edit the config file
Now you are ready for training. As we explain during the course, the Tf Object Detection API relies on **config files**. The config that we will use for this project is `pipeline.config`, which is the config for a SSD Resnet 50 640x640 model. You can learn more about the Single Shot Detector [here](https://arxiv.org/pdf/1512.02325.pdf).
First, let's download the [pretrained model](http://download.tensorflow.org/models/object_detection/tf2/20200711/ssd_resnet50_v1_fpn_640x640_coco17_tpu-8.tar.gz) and move it to `/home/workspace/experiments/pretrained_model/`.
We need to edit the config files to change the location of the training and validation files, as well as the location of the label_map file, pretrained weights. We also need to adjust the batch size. To do so, run the following:
python edit_config.py --train_dir /home/workspace/data/train/ --eval_dir /home/workspace/data/val/ --batch_size 2 --checkpoint /home/workspace/experiments/pretrained_model/ssd_resnet50_v1_fpn_640x640_coco17_tpu-8/checkpoint/ckpt-0 --label_map /home/workspace/experiments/label_map.pbtxt
A new config file has been created, `pipeline_new.config`.
### Training
You will now launch your very first experiment with the Tensorflow object detection API. Move the `pipeline_new.config` to the `/home/workspace/experiments/reference` folder. Now launch the training process:
* a training process:
python experiments/model_main_tf2.py --model_dir=experiments/reference/ --pipeline_config_path=experiments/reference/pipeline_new.config
Once the training is finished, launch the evaluation process:
* an evaluation process:
python experiments/model_main_tf2.py --model_dir=experiments/reference/ --pipeline_config_path=experiments/reference/pipeline_new.config --checkpoint_dir=experiments/reference/
**Note**: Both processes will display some Tensorflow warnings, which can be ignored. You may have to kill the evaluation script manually using
`CTRL+C`.
To monitor the training, you can launch a tensorboard instance by running `python -m tensorboard.main --logdir experiments/reference/`. You will report your findings in the writeup.
### Improve the performances
Most likely, this initial experiment did not yield optimal results. However, you can make multiple changes to the config file to improve this model. One obvious change consists in improving the data augmentation strategy. The [`preprocessor.proto`](https://github.com/tensorflow/models/blob/master/research/object_detection/protos/preprocessor.proto) file contains the different data augmentation method available in the Tf Object Detection API. To help you visualize these augmentations, we are providing a notebook: `Explore augmentations.ipynb`. Using this notebook, try different data augmentation combinations and select the one you think is optimal for our dataset. Justify your choices in the writeup.
Keep in mind that the following are also available:
* experiment with the optimizer: type of optimizer, learning rate, scheduler etc
* experiment with the architecture. The Tf Object Detection API [model zoo](https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/tf2_detection_zoo.md) offers many architectures. Keep in mind that the `pipeline.config` file is unique for each architecture and you will have to edit it.
**Important:** If you are working on the workspace, your storage is limited. You may to delete the checkpoints files after each experiment. You should however keep the `tf.events` files located in the `train` and `eval` folder of your experiments. You can also keep the `saved_model` folder to create your videos.
### Creating an animation
#### Export the trained model
Modify the arguments of the following function to adjust it to your models:
python experiments/exporter_main_v2.py --input_type image_tensor --pipeline_config_path experiments/reference/pipeline_new.config --trained_checkpoint_dir experiments/reference/ --output_directory experiments/reference/exported/
This should create a new folder `experiments/reference/exported/saved_model`. You can read more about the Tensorflow SavedModel format [here](https://www.tensorflow.org/guide/saved_model).
Finally, you can create a video of your model's inferences for any tf record file. To do so, run the following command (modify it to your files):
python inference_video.py --labelmap_path label_map.pbtxt --model_path experiments/reference/exported/saved_model --tf_record_path /data/waymo/testing/segment-12200383401366682847_2552_140_2572_140_with_camera_labels.tfrecord --config_path experiments/reference/pipeline_new.config --output_path animation.gif
## Submission Template
### Project overview
This section should contain a brief description of the project and what we are trying to achieve. Why is object detection such an important component of self driving car systems?
### Set up
This section should contain a brief description of the steps to follow to run the code for this repository.
### Dataset
#### Dataset analysis
This section should contain a quantitative and qualitative description of the dataset. It should include images, charts and other visualizations.
#### Cross validation
This section should detail the cross validation strategy and justify your approach.
### Training
#### Reference experiment
This section should detail the results of the reference experiment. It should includes training metrics and a detailed explanation of the algorithm's performances.
#### Improve on the reference
This section should highlight the different strategies you adopted to improve your model. It should contain relevant figures and details of your findings.
In the notebook Exploratory Data Analysis, The function display_images is implemented. The file plot the given batch in subplots with the given shape and figure size. The generated sub-plot is saved to a png file with the given file name. The bounding boxes are plotted with class dependent color.
Sample frames with ground truth bounding boxes for the training dataset:
Sample frames with ground truth bounding boxes for the validation dataset:
Sample frames with ground truth bounding boxes for the validation dataset. The ground truth bounding boxes in the provided test data seem to have scaling issues. That is the root cause of missing boxes. That is not a real issue in our use case, as these frames are only used for video creation.
The distribution of a random subset of the given training, validation and data set is shown in the following. As shown, the class distribution is significantly imbalanced.
Proposals to enhance the balance of object classes:
- Collect more recordings from cities with more dense cyclist and pedestrians (ex: Amsterdam, Bangalore, ).
- Augment the frames with dense pedestrians and cyclist more, so that we have more variants of these frames.
- Work more with synthetic data.
To ensure less over-fitting the The Following augmentations are included with their default values.
- RandomRGBtoGray: To avoid memorizing objects' color during training.
- RandomDistortColor: To avoid memorizing objects' color during training.
- RandomBlackPatches: To train the network on occlusion.
- RandomAdjustBrightness: To learn in different situations of lighting
- RandomAdjustContrast: For less sensitivity on image contrast
- RandomAdjustSaturation: For less sensitivity on color saturation
- RandomHorizontalFlip: The horizontal flip shall have no effect on the classification problem. (ex: right or left hand driving shall has no effect on classification)
- RandomCropImage: To have artificially more variations in perspective.
The notebook Explore augmentations is used to show a batch of the augmented frames.
It is observed that increasing the batch size did ensure much less loss value as well as less fluctuation compared to the reference training configuration. Thus, the batch size is increased from 2 to 16. Increasing the batch size more did raise memory issues in the workspace. If you are using local setup with more resources, I'd recommend even bigger batch size.
I'm presenting here the results of the second experiment, as it was more successful. As shown the validation loss is only slightly higher than that of training, which is a sign that the model did not over fit the training data. The exported videos are shown below.
