Update app.py

app.py

@@ -6,109 +6,29 @@ from decord import cpu, VideoReader, bridge
 from transformers import AutoModelForCausalLM, AutoTokenizer
 from transformers import BitsAndBytesConfig
 
+
 MODEL_PATH = "THUDM/cogvlm2-video-llama3-chat"
 DEVICE = 'cuda' if torch.cuda.is_available() else 'cpu'
 TORCH_TYPE = torch.bfloat16 if torch.cuda.is_available() and torch.cuda.get_device_capability()[0] >= 8 else torch.float16
 
-# Delay Reasons for Each Manufacturing Step
+
 DELAY_REASONS = {
-    "Step 1": ["Delay in Bead Insertion", "Lack of raw material"],
-    "Step 2": ["Inner Liner Adjustment by Technician", "Person rebuilding defective Tire Sections"],
-    "Step 3": ["Manual Adjustment in Ply1 apply", "Technician repairing defective Tire Sections"],
-    "Step 4": ["Delay in Bead set", "Lack of raw material"],
-    "Step 5": ["Delay in Turnup", "Lack of raw material"],
-    "Step 6": ["Person Repairing sidewall", "Person rebuilding defective Tire Sections"],
-    "Step 7": ["Delay in sidewall stitching", "Lack of raw material"],
-    "Step 8": ["No person available to load Carcass", "No person available to collect tire"]
+    "Step 1": ["Delay in Bead Insertion","Lack of raw material"],
+    "Step 2": ["Inner Liner Adjustment by Technician","Person rebuilding defective Tire Sections"],
+    "Step 3": ["Manual Adjustment in Ply1 apply","Technician repairing defective Tire Sections"],
+    "Step 4": ["Delay in Bead set","Lack of raw material"],
+    "Step 5": ["Delay in Turnup","Lack of raw material"],
+    "Step 6": ["Person Repairing sidewall","Person rebuilding defective Tire Sections"],
+    "Step 7": ["Delay in sidewall stitching","Lack of raw material"],
+    "Step 8": ["No person available to load Carcass","No person available to collect tire"]
 }
 
-def get_step_info(step_number):
-    """Returns detailed information about a manufacturing step."""
-    step_details = {
-        1: {
-            "Name": "Bead Insertion",
-            "Standard Time": "4 seconds",
-            "Video_substeps_expected": {
-                "0-1 second": "Machine starts bead insertion process.",
-                "1-3 seconds": "Beads are aligned and positioned.",
-                "3-4 seconds": "Final adjustment and confirmation of bead placement."
-            }
-        },
-        2: {
-            "Name": "Inner Liner Apply",
-            "Standard Time": "4 seconds",
-            "Video_substeps_expected": {
-                "0-1 second": "Machine applies the first layer of the liner.",
-                "1-3 seconds": "Technician checks alignment and adjusts if needed.",
-                "3-4 seconds": "Final inspection and confirmation."
-            }
-        },
-        3: {
-            "Name": "Ply1 Apply",
-            "Standard Time": "4 seconds",
-            "Video_substeps_expected": {
-                "0-2 seconds": "First ply is loaded onto the machine.",
-                "2-4 seconds": "Technician inspects and adjusts ply placement."
-            }
-        },
-        4: {
-            "Name": "Bead Set",
-            "Standard Time": "8 seconds",
-            "Video_substeps_expected": {
-                "0-3 seconds": "Bead is positioned and pre-set.",
-                "3-6 seconds": "Machine secures the bead in place.",
-                "6-8 seconds": "Technician confirms the bead alignment."
-            }
-        },
-        5: {
-            "Name": "Turnup",
-            "Standard Time": "4 seconds",
-            "Video_substeps_expected": {
-                "0-2 seconds": "Turnup process begins with machine handling.",
-                "2-4 seconds": "Technician inspects the turnup and makes adjustments if necessary."
-            }
-        },
-        6: {
-            "Name": "Sidewall Apply",
-            "Standard Time": "14 seconds",
-            "Video_substeps_expected": {
-                "0-5 seconds": "Sidewall material is positioned by the machine.",
-                "5-10 seconds": "Technician checks for alignment and begins application.",
-                "10-14 seconds": "Final adjustments and confirmation of sidewall placement."
-            }
-        },
-        7: {
-            "Name": "Sidewall Stitching",
-            "Standard Time": "5 seconds",
-            "Video_substeps_expected": {
-                "0-2 seconds": "Stitching process begins automatically.",
-                "2-4 seconds": "Technician inspects stitching for any irregularities.",
-                "4-5 seconds": "Machine completes stitching process."
-            }
-        },
-        8: {
-            "Name": "Carcass Unload",
-            "Standard Time": "7 seconds",
-            "Video_substeps_expected": {
-                "0-3 seconds": "Technician unloads(removes) carcass(tire) from the machine."
-            },
-            "Potential_Delay_reasons": [
-                "Person not available in time(in 3 sec) to remove carcass.",
-                "Person is doing bead(ring) insertion before carcass unload causing unload to be delayed by more than 3 sec"
-            ]
-        }
-    }
-    
-    return step_details.get(step_number, {"Error": "Invalid step number. Please provide a valid step number."})
-
-
-
 def load_video(video_data, strategy='chat'):
     """Loads and processes video data into a format suitable for model input."""
     bridge.set_bridge('torch')
     num_frames = 24
     
-    if isinstance(video_data, str):
+    if isinstance(video_data, str): 
         decord_vr = VideoReader(video_data, ctx=cpu(0))
     else:  
         decord_vr = VideoReader(io.BytesIO(video_data), ctx=cpu(0))
@@ -171,7 +91,7 @@ def predict(prompt, video_data, temperature, model, tokenizer):
     gen_kwargs = {
         "max_new_tokens": 2048,
         "pad_token_id": 128002,
-        "top_k": 3,
+        "top_k": 1,
         "do_sample": False,
         "top_p": 0.1,
         "temperature": temperature,
@@ -184,62 +104,38 @@ def predict(prompt, video_data, temperature, model, tokenizer):
     
     return response
 
-def get_analysis_prompt(step_number):
+def get_analysis_prompt(step_number, possible_reasons):
     """Constructs the prompt for analyzing delay reasons based on the selected step."""
-    step_info = get_step_info(step_number)
-    
-    if "Error" in step_info:
-        return step_info["Error"]
-    
-    step_name = step_info["Name"]
-    standard_time = step_info["Standard Time"]
-    analysis = step_info["Analysis"]
-
-    return f"""
-You are an AI expert system specialized in analyzing manufacturing processes and identifying production delays in tire manufacturing. Your role is to accurately classify delay reasons based on visual evidence from production line footage.
+    return f"""You are an AI expert system specialized in analyzing manufacturing processes and identifying production delays in tire manufacturing. Your role is to accurately classify delay reasons based on visual evidence from production line footage.
 Task Context:
-You are analyzing video footage from Step {step_number} of a tire manufacturing process where a delay has been detected. The step is called {step_name}, and its standard time is {standard_time}.
+You are analyzing video footage from Step {step_number} of a tire manufacturing process where a delay has been detected. Your task is to determine the most likely cause of the delay from the following possible reasons:
+{', '.join(possible_reasons)}
 Required Analysis:
-Carefully observe the video for visual cues indicating production interruption.
-- If no person is visible in any of the frames, the reason probably might be due to their absence.
-- If a person is visible in the video and is observed touching and modifying the layers of the tire, it indicates an issue with tire patching, and the person might be repairing it.
-- Compare observed evidence against the following possible delay reasons: 
-  - {analysis}
-Following are the subactivities needs to happen in this step.
-
-{get_step_info(step_number)}
-
-Important:Please provide your output in the following format:
-
-Output_Format = {
-     ["Delay in Bead Insertion", "Lack of raw material"],
-     ["Inner Liner Adjustment by Technician", "Person rebuilding defective Tire Sections"],
-     ["Manual Adjustment in Ply1 Apply", "Technician repairing defective Tire Sections"],
-     ["Delay in Bead Set", "Lack of raw material"],
-     ["Delay in Turnup", "Lack of raw material"],
-     ["Person Repairing Sidewall", "Person rebuilding defective Tire Sections"],
-     ["Delay in Sidewall Stitching", "Lack of raw material"],
-     ["No person available to load Carcass", "No person available to collect tire"]
-}
-1. **Selected Reason:** [State the most likely reason from the given options]
-2. **Visual Evidence:** [Describe specific visual cues that support your selection]
-3. **Reasoning:** [Explain why this reason best matches the observed evidence]
-4. **Alternative Analysis:** [Brief explanation of why other possible reasons are less likely]
-Base your analysis solely on visual evidence from the video. Focus on concrete, observable details rather than assumptions. Clearly state if no person or specific activity is observed.
-"""
-
-
-
+ Carefully observe the video for visual cues indicating production interruption.
+ If no person is visible in any of the frames, the reason probably might be due to his absence.
+ If a person is visible in the video and is observed touching and modifying the layers of the tire, it means there is a issue with tyre being patched hence he is repairing it.
+ Compare observed evidence against each possible delay reason.
+ Select the most likely reason based on visual evidence.
+Please provide your analysis in the following format:
+1. Selected Reason: [State the most likely reason from the given options]
+2. Visual Evidence: [Describe specific visual cues that support your selection]
+3. Reasoning: [Explain why this reason best matches the observed evidence]
+4. Alternative Analysis: [Brief explanation of why other possible reasons are less likely]
+Important: Base your analysis solely on visual evidence from the video. Focus on concrete, observable details rather than assumptions. Clearly state if no person or specific activity is observed."""
+
+
+# Load model globally
 model, tokenizer = load_model()
 
 def inference(video, step_number):
-    """Analyzes video to predict possible issues based on the manufacturing step."""
+    """Analyzes video to predict the most likely cause of delay in the selected manufacturing step."""
     try:
         if not video:
             return "Please upload a video first."
         
-        prompt = get_analysis_prompt(step_number)
-        temperature = 0.3
+        possible_reasons = DELAY_REASONS[step_number]
+        prompt = get_analysis_prompt(step_number, possible_reasons)
+        temperature = 0.8
         response = predict(prompt, video, temperature, model, tokenizer)
         
         return response
@@ -247,34 +143,38 @@ def inference(video, step_number):
         return f"An error occurred during analysis: {str(e)}"
 
 def create_interface():
-    """Creates the Gradio interface for the Manufacturing Analysis System."""
+    """Creates the Gradio interface for the Manufacturing Delay Analysis System with examples."""
     with gr.Blocks() as demo:
         gr.Markdown("""
-        # Manufacturing Analysis System
+        # Manufacturing Delay Analysis System
         Upload a video of the manufacturing step and select the step number. 
-        The system will analyze the video and provide observations.
+        The system will analyze the video and determine the most likely cause of delay.
         """)
         
         with gr.Row():
             with gr.Column():
                 video = gr.Video(label="Upload Manufacturing Video", sources=["upload"])
                 step_number = gr.Dropdown(
-                    choices=[f"Step {i}" for i in range(1, 9)],
+                    choices=list(DELAY_REASONS.keys()),
                     label="Manufacturing Step"
                 )
-                analyze_btn = gr.Button("Analyze", variant="primary")
+                analyze_btn = gr.Button("Analyze Delay", variant="primary")
             
             with gr.Column():
                 output = gr.Textbox(label="Analysis Result", lines=10)
         
+        # Add examples
+        examples = [
+            ["7838_step2_2_eval.mp4", "Step 2"],
+            ["7838_step6_2_eval.mp4", "Step 6"],
+            ["7838_step8_1_eval.mp4", "Step 8"],
+            ["7993_step6_3_eval.mp4", "Step 6"],
+            ["7993_step8_3_eval.mp4", "Step 8"]
+            
+        ]
+        
         gr.Examples(
-            examples=[
-                ["7838_step2_2_eval.mp4", "Step 2"],
-                ["7838_step6_2_eval.mp4", "Step 6"],
-                ["7838_step8_1_eval.mp4", "Step 8"],
-                ["7993_step6_3_eval.mp4", "Step 6"],
-                ["7993_step8_3_eval.mp4", "Step 8"]    
-            ],
+            examples=examples,
             inputs=[video, step_number],
             cache_examples=False
         )