feat: add prompts

prompts/built_elements.txt

@@ -0,0 +1,17 @@
+Identify any man-made structures in the image, including benches, kiosques, pergolas, playground equipment, paved paths, trash bins, street lamps, or information signs.
+
+Do not use any external data sources.
+If the image does not belong to any of the categories, return 'Other'.
+If the image is not clear enough to make a decision, return 'Not clear'.
+If the image is not relevant to the task, return 'Irrelevant'.
+If the image can be classified into multiple categories, return the most relevant one.
+
+Results must be consistent across multiple runs of the model on the same image.
+Results must be JSON serializable.
+
+**Output format:**  
+```
+{
+    "elements": ["benches", "kiosques", "pergolas", "..."]
+}
+```

prompts/fauna_identification.txt

@@ -0,0 +1,32 @@
+Analyze the image and identify any animals present. Classify detected fauna into one of the following categories:
+- Birds
+- Insects
+- Rodents
+- Foxes, deer, or other large mammals
+- Amphibians
+
+For each detected animal, provide its type and count.
+
+Do not use any external data sources.
+If no fauna is detected, return an empty list.
+If the image is not clear enough to make a decision, return 'Not clear'.
+If the image is not relevant to the task, return 'Irrelevant'.
+
+Results must be consistent across multiple runs of the model on the same image.
+Results must be JSON serializable.
+
+**Output format:**  
+```
+{
+    "fauna": [
+        {
+            "type": "Birds",
+            "count": 3
+        },
+        {
+            "type": "Insects",
+            "count": 5
+        }
+    ]
+}
+```

prompts/general_classification.txt

@@ -0,0 +1,20 @@
+Analyze the image and detect any man-made structures. Identify and list the following elements if they are present:
+- Benches, kiosques, pergolas
+- Playgrounds (swings, slides)
+- Paved or asphalted paths
+- Trash bins, street lamps, and information signs
+
+Do not use any external data sources.
+If no built elements are detected, return an empty list.
+If the image is not clear enough to make a decision, return 'Not clear'.
+If the image is not relevant to the task, return 'Irrelevant'.
+
+Results must be consistent across multiple runs of the model on the same image.
+Results must be JSON serializable.
+
+**Output format:**  
+```
+{
+    "classification": "Espaces verts artificialisés"
+}
+```

prompts/human_activity.txt

@@ -0,0 +1,35 @@
+Analyze the image and detect any traces of human activity. Identify and categorize them as:
+- Picnic remains
+- Trash or debris
+- Footprints or worn paths
+- Abandoned objects
+
+For each detected trace, provide its type.
+
+Do not use any external data sources.
+If no traces of human activity are detected, return an empty list.
+If the image is not clear enough to make a decision, return 'Not clear'.
+If the image is not relevant to the task, return 'Irrelevant'.
+
+Results must be consistent across multiple runs of the model on the same image.
+Results must be JSON serializable.
+
+**Output format:**  
+```
+{
+  "traces": [
+    {
+      "type": "Picnic remains"
+    },
+    {
+      "type": "Trash or debris"
+    },
+    {
+      "type": "Footprints or worn paths"
+    },
+    {
+      "type": "Abandoned objects"
+    }
+  ]
+}
+```

prompts/human_detection.txt

@@ -0,0 +1,23 @@
+Analyze the image and detect human presence. Identify and categorize detected individuals as:
+- Pedestrians
+- Cyclists
+- People with strollers
+
+Count the number of individuals for each category.
+
+Do not use any external data sources.
+If no human presence is detected, return an empty object.
+If the image is not clear enough to make a decision, return 'Not clear'.
+If the image is not relevant to the task, return 'Irrelevant'.
+
+Results must be consistent across multiple runs of the model on the same image.
+Results must be JSON serializable.
+
+**Output format:**  
+````
+{
+    "pedestrians": 0,
+    "cyclists": 0,
+    "people_with_strollers": 0
+}
+```

prompts/vegetation_detection.txt

@@ -0,0 +1,19 @@
+Analyze the image and identify the types of vegetation present. Categorize vegetation into one or more of the following types:
+- Maintained (e.g., trimmed hedges, flower beds, decorative shrubs, lawns)
+- Mixed (e.g., tall grass, prairies, untrimmed shrubs, scattered trees)
+- Dense and wild (e.g., forests, undergrowth, thick bushes, wetlands)
+
+Do not use any external data sources.
+If no vegetation is detected, return an empty list.
+If the image is not clear enough to make a decision, return 'Not clear'.
+If the image is not relevant to the task, return 'Irrelevant'.
+
+Results must be consistent across multiple runs of the model on the same image.
+Results must be JSON serializable.
+
+**Output format:**  
+```
+{
+    "vegetation": [ "Maintained", "Mixed" ]
+}
+```

prompts/water_elements.txt

@@ -0,0 +1,29 @@
+Analyze the image and detect any water elements. Identify and classify them as:
+- Artificial (e.g., fountains, artificial ponds)
+- Natural (e.g., rivers, lakes, waterfalls)
+
+For each detected water element, provide its type and classification.
+
+Do not use any external data sources.
+If no water elements are detected, return an empty list.
+If the image is not clear enough to make a decision, return 'Not clear'.
+If the image is not relevant to the task, return 'Irrelevant'.
+
+Results must be consistent across multiple runs of the model on the same image.
+Results must be JSON serializable.
+
+**Output format:**  
+```
+{
+    "water_elements": [
+        {
+            "type": "Artificial",
+            "classification": "Fountain"
+        },
+        {
+            "type": "Natural",
+            "classification": "River"
+        }
+    ]
+}
+```