From f67bc22a4d409e7d0d1ca45d216a2bbdb0988ce5 Mon Sep 17 00:00:00 2001 From: Byungheong Jeong <byungheon.jeong@gmail.com> Date: Thu, 13 Feb 2020 17:39:17 -0800 Subject: [PATCH] Update README.md --- README.md | 42 +++++++++++++++++++++--------------------- 1 file changed, 21 insertions(+), 21 deletions(-) diff --git a/README.md b/README.md index 6b03b8f..3bb64b4 100644 --- a/README.md +++ b/README.md @@ -44,21 +44,21 @@ If you are planning to use this implementation on another Nautilus namespace, th 1. Changing namespace address <br /> <br /> - <br /> + <br /> **Change the name and the namespace entries to the current working namespace and a suitable name** 2. Change the resource requests <br /> <br /> - <br /> + <br /> **Change the numbers to suit the task** 3. Mount volumne <br /><br /> - <br /> + <br /> **Very important for crash-resistance. I highly recommend saving all work onto mounted directory** 4. Choose GPU type <br /><br /> - <br /> + <br /> If doing intensive training, choose larger/more expensive GPUs ## Using the Components @@ -66,25 +66,25 @@ If doing intensive training, choose larger/more expensive GPUs ### Starting the development and accessing jupyter notebook 1. Go into kerasDeloyment.yaml file 2. Choose the RAW file format <br /> - <br /> + <br /> 3. copy url of RAW file <br /> - <br /> + <br /> 4. execute yaml file on nautilius namespace <br /> - + 5. exec into nautilus pod <br /> - + 6. Navigate to /userdata/kerasData and Start Jupyter Notebook <br /><br /> -<br /> +<br /> **Note: The port number choice does not matter, as long as there are not other processes running on that port. If a port is already in use, jupyter will automatically assign another port. Make sure to match the port number in the next step** <br /> <br /> -<br /> +<br /> _What happens when a wrong port is chosen_ <br /> 7. Go to your computer terminal and start port-forward, matching the port in the pod <br /> -<br /> +<br /> 8. Go to the localhost address<br /> <br /> @@ -92,8 +92,8 @@ _What happens when a wrong port is chosen_ <br /> 9. Test for keras Create a new notebook or use the ClassificationExample.ipynb file - Run the following tests <br /> - <br /><br /> -<!-- <br /><br /> --> + <br /><br /> +<!-- <br /><br /> --> **_Make sure that the outputs return True or some name._**<br /> **You are now ready to use Keras on a jupyter notebook hosted on Kubernetes** @@ -101,11 +101,11 @@ Create a new notebook or use the ClassificationExample.ipynb file #### EXTREMELY IMPORTANT! In order to prevent Keras from assigning too much GPU memory and stalling training efforts later on, run this: - <br /> + <br /> If you see an error, shutdown the network server and try again <br /> -<br/> +<br/> If you see nvidia-smi memory allocation at 0/- you have suceeded in reseting the GPU <br /> -<br /> +<br /> Please refer to [Keras Documentation](https://keras.io/) for instructions and information @@ -117,9 +117,9 @@ I used the notebook for the following: ## Using the Fire-Classification training 1. Write the network using Keras layers <br /> - <br /> <br /> + <br /> <br /> 2. Set the paths <br /> - <br /> + <br /> The following must be set - FIRE_PATH = Path of the directory with the fire images - Non_FIRE_PATH = Path of the directory with images without fire @@ -143,9 +143,9 @@ More information is availbe here [pyimagesearch](https://www.pyimagesearch.com/2 Finally, fill out the INIT_LR from what you learned from above <br /> 7. Train <br /> - <br /> -8. Get results - <br /> + <br /> +8. Get results <br /> + <br /> You will find the accuracy measures in the table. Find the model in fire_detection.model -- GitLab