first commit

app.py

@@ -0,0 +1,85 @@
+from typing import Any, List
+import asyncio
+import json
+import logging
+import os
+
+from albumentations import Compose, LongestMaxSize, Normalize, PadIfNeeded
+from albumentations.pytorch import ToTensorV2
+import cv2
+import streamlit as st
+import torch
+import PIL
+import numpy as np
+
+class ClassifyModel:
+    def __init__(self):
+        self.model = None
+        self.class2tag = None
+        self.tag2class = None
+        self.transform = None
+
+    def load(self, path="/model"):
+        image_size = 512
+        self.transform = Compose(
+            [
+                LongestMaxSize(max_size=image_size),
+                PadIfNeeded(image_size, image_size, border_mode=cv2.BORDER_CONSTANT),
+                Normalize((0.485, 0.456, 0.406), (0.229, 0.224, 0.225), always_apply=True),
+                ToTensorV2()
+            ]
+        )
+        self.model = torch.jit.load("model_healthy_bot.pth")
+        with open("tag2class_healthy_bot.json") as fin:
+            self.tag2class = json.load(fin)
+            self.class2tag = {v: k for k, v in self.tag2class.items()}
+            logging.debug(f"class2tag: {self.class2tag}")
+
+    def predict(self, *imgs) -> List[str]:
+        logging.debug(f"batch size: {len(imgs)}")
+        input_ts = [self.transform(image=img)["image"] for img in imgs]
+        input_t = torch.stack(input_ts)
+        logging.debug(f"input_t: {input_t.shape}")
+        output_ts = self.model(input_t)
+        activation_fn = torch.nn.__dict__['Sigmoid']()
+        output_ts = activation_fn(output_ts)
+        labels = list(self.tag2class.keys())
+        logging.debug(f"output_ts: {output_ts.shape}")
+        #logging.debug(f"output_pb: {output_pb}")
+        res = []
+        trh = 0.5
+        for output_t in output_ts:
+            logit = (output_t > trh).long()
+            if logit[0] and any([*logit[1:3], *logit[4:]]): 
+                output_t[0] = 0
+            indices = (output_t > trh).nonzero(as_tuple=True)[0]
+            prob = output_t[indices].tolist()
+            tag  = [labels[i] for i in indices.tolist()]
+            res_dict = dict(zip(
+                         list(self.tag2class.keys()),list(output_t.numpy())
+                       ))
+            logging.debug(f"all results: {res_dict}")
+            logging.debug(f"prob: {prob}")
+            logging.debug(f"result: {tag}")
+            res.append((tag,prob,res_dict))
+        return res
+
+m = ClassifyModel()
+m.load()
+
+st.sidebar.title("About")
+
+st.sidebar.info(
+    "This application identifies the crop health in the picture.")
+
+
+st.title('Wheat Rust Identification')
+model = load_tf_model(model_path)
+st.write("Upload an image.")
+uploaded_file = st.file_uploader("")
+
+if uploaded_file is not None:
+    image = Image.open(uploaded_file)
+    img = np.array(image)
+    result = m.predict(img)
+    st.write(f"I think this has {result[0][0]}(confidence: {round(result[0][1],2)})")

model_healthy_bot.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d705f89edd648d4f6f55917cac6205d1eb8851b0d1afe872d5c3386df679a871
+size 21884854

tag2class_healthy_bot.json

@@ -0,0 +1,9 @@
+{
+ "healthy": 0,
+ "leaf rust": 1,
+ "powdery mildew": 2,
+ "seedlings": 3,
+ "septoria": 4,
+ "stem rust": 5,
+ "yellow rust": 6
+}