Update README.md

README.md

@@ -15,7 +15,7 @@ Now project is base on [YOLOv5](https://github.com/ultralytics/yolov5) repo but
 - For tagging i use [Label Studio](https://labelstud.io/)
 
 ### Project pipeline
-insert pipeline image
+![image](https://github.com/PawelKinczyk/AECVision/assets/96824698/b869920b-639a-41b4-b434-97d19bdc7f4c)
 
 ### Available models
 
@@ -34,12 +34,18 @@ insert pipeline image
 The model was trained on clear plans and with annotation but remember that the best results you get without many symbols on construction plans. 
 
 ### Detection
+#### Confidence = 0.5
+![Za -5-Projekt-architektoniczno-budowlany pdf2_2_3](https://github.com/PawelKinczyk/AECVision/assets/96824698/f7b43709-e9e5-4b16-9723-70306e202125)
+#### Confidence = 0.8
+![Za -5-Projekt-architektoniczno-budowlany pdf2_2_31](https://github.com/PawelKinczyk/AECVision/assets/96824698/de853059-628b-4474-831e-765c59835d64)
 
-images
 
 ### Screen detection
+By using screen_detection.py
+
+https://github.com/PawelKinczyk/AECVision/assets/96824698/f311eab2-11e9-483b-9a99-11136c1eaf29
+
 
-dddd past gif
 ## Project roadmap
 
 - [ ]  Collect and tag more images (up to 500)
@@ -54,7 +60,31 @@ dddd past gif
 ## Summary & Problems
 
 ### First training
+#### Data description
+Walls are overrepresented because this is normal quantities in architectural plans.
+
+![image](https://github.com/PawelKinczyk/AECVision/assets/96824698/e515fdd1-bf00-40f6-9216-5d6eb6fd6c7f)
+
+Maybe in future, i need to create more plans with other categories to improve other classes detection.
+
+#### Learning process
+##### Model has steadily improved up to the 272nd epoch and probably has further potential.
+
+![image](https://github.com/PawelKinczyk/AECVision/assets/96824698/83dbea04-a7cd-4380-ae17-2158114dec9a)
+
+##### On the validation set, an increase in "obj_loss" was noted during the learning process.
+
+![image](https://github.com/PawelKinczyk/AECVision/assets/96824698/cdcf25ed-8ea4-4ab1-8f2e-d3655176fdf7)
+
+##### Confussion matrix show that all classes except "background" are correctly identified.
+
+![image](https://github.com/PawelKinczyk/AECVision/assets/96824698/0f917feb-e89d-4fe7-9df9-4eb1d18c59fb)
+
+##### If you want to dig into training results see train_results or write to me.
 
+#### Problems
+- A small number of certain classes
+- Conversion from pdf (default construction plans format) to jpg results in a loss of image quality
 
 ## Contributing