What happened when I was doing Server/Machine Learning

the first thing I learned since I came to People space was Fastai. I think I had a hard time setting the path to the folder while studying. my file path were so different, with fastai lesson2 I have to designate it directly through the SDK prompt. At first, I misinterpreted the script that Patrick sent me, so I had a hard time getting a new machine learning model, but I realized that I didn't have to update it every time. I've never done a server before, so I didn't know what routing was or what the frame work or template engine was. but I could slowly find out when I asked Christian or Patrick.

 

The server was fine, but there were many problems caused by studying machine learning models. Collecting disease data sets for special plants, the class designation itself was a reckless situation, as there were not many photos and not many people with expertise. it was not possible to do multi-classification.

So, after much thought, I decided to divide it into a model that identifies disease and a model that identifies plant species.

The data on the disease was particularly scarce, so I had to find diseases that occurred in common plants and

I could find sunburn,bacterial spot,powdery mildew, brown spot, black rot and so on. and FInally collecting 400 data set each classes.

(I can find awesome dataset from github, https://medium.com/analytics-vidhya/plants-doctor-detecting-diseases-in-plants-e5c44c569bdb there's diseases of plant about normal plant like tomato, grape etc. and I use it)

I was able to create machine learning model files and write code that I used by importing modules from servers.

But there was another problem. As I wanted to check the image and run(collecting data part), it became difficult to upload it through sdk as I received the jpg file instead of csv where url was stored. I had to use a bucket to upload folders(google cloud platform), but the Wi-Fi was so slow that I had to automatically cancel my photos while uploading them.

So I downloaded fastai directly in the local area and tried again, but got a message to buy my ram with runtime error (I know my laptop is bad...THE LAPTOP THAT EVEN JUPYTER ACKNOWLEDGE)

 

in the end, the machine learning model could be turned through the colab. next time, I try to write with more diverse problems.