feat: created new tasks with warehouse simulation images

Magdalena Kotynia · Magdalena Kotynia · commit 2a59c06100ff · 2025-08-12T10:07:09.000+02:00
diff --git a/.gitattributes b/.gitattributes
@@ -0,0 +1,9 @@
+*.png filter=lfs diff=lfs merge=lfs -text
+*.jpg filter=lfs diff=lfs merge=lfs -text
+*.jpeg filter=lfs diff=lfs merge=lfs -text
+*.gif filter=lfs diff=lfs merge=lfs -text
+*.webp filter=lfs diff=lfs merge=lfs -text
+*.bmp filter=lfs diff=lfs merge=lfs -text
+*.tif filter=lfs diff=lfs merge=lfs -text
+*.tiff filter=lfs diff=lfs merge=lfs -text
+*.svg filter=lfs diff=lfs merge=lfs -text
diff --git a/src/rai_bench/rai_bench/vlm_benchmark/predefined/images/damaged_boxes.png b/src/rai_bench/rai_bench/vlm_benchmark/predefined/images/damaged_boxes.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:edd337c10f0c662bfedbd6a5137d6e52decc7d2937134b222d76f7f34f588228
+size 2506357
diff --git a/src/rai_bench/rai_bench/vlm_benchmark/predefined/images/fallen_barrels_1.png b/src/rai_bench/rai_bench/vlm_benchmark/predefined/images/fallen_barrels_1.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ece7ce624db2cbe081eacd503de97501b303a2c38c6d2902fefdd577d1046ebd
+size 497598
diff --git a/src/rai_bench/rai_bench/vlm_benchmark/predefined/images/fallen_barrels_2.png b/src/rai_bench/rai_bench/vlm_benchmark/predefined/images/fallen_barrels_2.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:389dc7b475f338bd4bce0aa477127e4e601075d2e5a545cdc414f506c78e0a0f
+size 1139409
diff --git a/src/rai_bench/rai_bench/vlm_benchmark/predefined/images/fallen_barrels_3.png b/src/rai_bench/rai_bench/vlm_benchmark/predefined/images/fallen_barrels_3.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9fd24939958439f2d2b193ef98306382b0eb4700be22ddf964dacf422043eccc
+size 232378
diff --git a/src/rai_bench/rai_bench/vlm_benchmark/predefined/images/fallen_human_1.png b/src/rai_bench/rai_bench/vlm_benchmark/predefined/images/fallen_human_1.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d515a1ebf1849ccdc7d3467c5eaf4483d8300fb672a56a928ea074b55ab8f384
+size 510974
diff --git a/src/rai_bench/rai_bench/vlm_benchmark/predefined/images/fallen_human_2.png b/src/rai_bench/rai_bench/vlm_benchmark/predefined/images/fallen_human_2.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:16a84ef5c9a4ec76c22b665c97be9c17610dee314bd9cfcba087991a8fcc6415
+size 1195412
diff --git a/src/rai_bench/rai_bench/vlm_benchmark/predefined/images/fallen_human_3.png b/src/rai_bench/rai_bench/vlm_benchmark/predefined/images/fallen_human_3.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:44cd817356eb6fc8bd90e4f4ad04a31f8ba2a7c441f8fa2433c24188bd42969d
+size 222920
diff --git a/src/rai_bench/rai_bench/vlm_benchmark/predefined/images/gauge_level.png b/src/rai_bench/rai_bench/vlm_benchmark/predefined/images/gauge_level.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6d158ca524cb0faa8a3bed64ae46f05f488a9597abd332a9810d800757aaeddc
+size 441519
diff --git a/src/rai_bench/rai_bench/vlm_benchmark/predefined/images/woman_wo_helmet_1.png b/src/rai_bench/rai_bench/vlm_benchmark/predefined/images/woman_wo_helmet_1.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:923a9ab71e68715d7c927e5c218ba406cc47f37e6145ccf4d3dc19512d040fe9
+size 6205972
diff --git a/src/rai_bench/rai_bench/vlm_benchmark/predefined/images/woman_wo_helmet_2.png b/src/rai_bench/rai_bench/vlm_benchmark/predefined/images/woman_wo_helmet_2.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:09ea1c58a154bc565bb74f2c91f2cec6c95613a5772c65eb99cdd86bf4982bed
+size 679496
diff --git a/src/rai_bench/rai_bench/vlm_benchmark/predefined/tasks.py b/src/rai_bench/rai_bench/vlm_benchmark/predefined/tasks.py
@@ -17,7 +17,14 @@
 from pydantic import BaseModel
 
 from rai_bench.vlm_benchmark.interfaces import ImageReasoningTask
-from rai_bench.vlm_benchmark.tasks.tasks import BoolImageTask, BoolImageTaskInput
+from rai_bench.vlm_benchmark.tasks.tasks import (
+    BoolImageTask,
+    BoolImageTaskInput,
+    MultipleChoiceImageTask,
+    MultipleChoiceImageTaskInput,
+    QuantityImageTask,
+    QuantityImageTaskInput,
+)
 
 IMG_PATH = "src/rai_bench/rai_bench/vlm_benchmark/predefined/images/"
 true_response_inputs: List[BoolImageTaskInput] = [
@@ -56,7 +63,38 @@
         images_paths=[IMG_PATH + "image_7.jpg"],
         expected_answer=True,
     ),
+    BoolImageTaskInput(
+        question="Do you see something anomalous in the image?",
+        images_paths=[IMG_PATH + "fallen_human_3.png"],
+        expected_answer=True,
+    ),
+    BoolImageTaskInput(
+        question="Do you see something anomalous in the image?",
+        images_paths=[IMG_PATH + "fallen_human_2.png"],
+        expected_answer=True,
+    ),
+    BoolImageTaskInput(
+        question="Do you see something anomalous in the image?",
+        images_paths=[IMG_PATH + "fallen_human_1.png"],
+        expected_answer=True,
+    ),
+    BoolImageTaskInput(
+        question="Is gauge level high?",
+        images_paths=[IMG_PATH + "gauge_level.png"],
+        expected_answer=True,
+    ),
+    BoolImageTaskInput(
+        question="Is worker wearing a vest?",
+        images_paths=[IMG_PATH + "woman_wo_helmet_2.png"],
+        expected_answer=True,
+    ),
+    BoolImageTaskInput(
+        question="Is any person in the image?",
+        images_paths=[IMG_PATH + "fallen_barrels_1.png"],
+        expected_answer=True,
+    ),
 ]
+
 false_response_inputs: List[BoolImageTaskInput] = [
     BoolImageTaskInput(
         question="Is the door open?",
@@ -93,6 +131,237 @@
         images_paths=[IMG_PATH + "image_7.jpg"],
         expected_answer=False,
     ),
+    BoolImageTaskInput(
+        question="Is worker wearing a helmet?",
+        images_paths=[IMG_PATH + "woman_wo_helmet_1.png"],
+        expected_answer=False,
+    ),
+    BoolImageTaskInput(
+        question="Is worker wearing a helmet?",
+        images_paths=[IMG_PATH + "woman_wo_helmet_2.png"],
+        expected_answer=False,
+    ),
+    BoolImageTaskInput(
+        question="Is gauge level low?",
+        images_paths=[IMG_PATH + "gauge_level.png"],
+        expected_answer=False,
+    ),
+]
+
+quantity_response_inputs: List[QuantityImageTaskInput] = [
+    QuantityImageTaskInput(
+        question="How many barrels are in the image?",
+        images_paths=[IMG_PATH + "fallen_barrels_1.png"],
+        expected_answer=6,
+    ),
+    QuantityImageTaskInput(
+        question="How many fallen barrels are in the image?",
+        images_paths=[IMG_PATH + "fallen_barrels_1.png"],
+        expected_answer=2,
+    ),
+    QuantityImageTaskInput(
+        question="How many fallen barrels are in the image?",
+        images_paths=[IMG_PATH + "fallen_barrels_2.png"],
+        expected_answer=1,
+    ),
+    QuantityImageTaskInput(
+        question="How many fallen barrels are in the image?",
+        images_paths=[IMG_PATH + "fallen_barrels_3.png"],
+        expected_answer=1,
+    ),
+    QuantityImageTaskInput(
+        question="How many barrels are in the image?",
+        images_paths=[IMG_PATH + "fallen_barrels_3.png"],
+        expected_answer=6,
+    ),
+    QuantityImageTaskInput(
+        question="How many people are in the image?",
+        images_paths=[IMG_PATH + "fallen_barrels_2.png"],
+        expected_answer=0,
+    ),
+    QuantityImageTaskInput(
+        question="How many people are in the image?",
+        images_paths=[IMG_PATH + "fallen_human_1.png"],
+        expected_answer=1,
+    ),
+    QuantityImageTaskInput(
+        question="How many people are in the image?",
+        images_paths=[IMG_PATH + "fallen_human_2.png"],
+        expected_answer=1,
+    ),
+    QuantityImageTaskInput(
+        question="How many people are in the image?",
+        images_paths=[IMG_PATH + "fallen_human_3.png"],
+        expected_answer=1,
+    ),
+    QuantityImageTaskInput(
+        question="How many barrels are in the image?",
+        images_paths=[IMG_PATH + "gauge_level.png"],
+        expected_answer=4,
+    ),
+    QuantityImageTaskInput(
+        question="How many damaged boxes are in the image?",
+        images_paths=[IMG_PATH + "damaged_boxes.png"],
+        expected_answer=2,
+    ),
+    QuantityImageTaskInput(
+        question="How many pictures are on the wall?",
+        images_paths=[IMG_PATH + "image_3.jpg"],
+        expected_answer=3,
+    ),
+    QuantityImageTaskInput(
+        question="How many pictures are on the wall?",
+        images_paths=[IMG_PATH + "image_4.jpg"],
+        expected_answer=3,
+    ),
+    QuantityImageTaskInput(
+        question="How many pictures are on the wall?",
+        images_paths=[IMG_PATH + "image_6.jpg"],
+        expected_answer=2,
+    ),
+    QuantityImageTaskInput(
+        question="How many windows are visible in the image?",
+        images_paths=[IMG_PATH + "image_2.jpg"],
+        expected_answer=2,
+    ),
+]
+
+multiple_choice_response_inputs: List[MultipleChoiceImageTaskInput] = [
+    MultipleChoiceImageTaskInput(
+        question="What is in the image?",
+        images_paths=[IMG_PATH + "fallen_barrels_2.png"],
+        options=[
+            "sofa",
+            "gauge",
+            "bed",
+            "armchair",
+            "barrels",
+            "lamp with a shade",
+            "door",
+            "boxes",
+            "human",
+            "desk",
+            "plant",
+            "roll container",
+            "dresser",
+        ],
+        expected_answer=["barrels", "boxes", "roll container"],
+    ),
+    MultipleChoiceImageTaskInput(
+        question="What is in the image?",
+        images_paths=[IMG_PATH + "fallen_human_1.png"],
+        options=[
+            "sofa",
+            "gauge",
+            "bed",
+            "armchair",
+            "barrels",
+            "lamp with a shade",
+            "door",
+            "boxes",
+            "human",
+            "desk",
+            "plant",
+            "roll container",
+            "dresser",
+        ],
+        expected_answer=["human", "boxes"],
+    ),
+    MultipleChoiceImageTaskInput(
+        question="What is in the image?",
+        images_paths=[IMG_PATH + "gauge_level.png"],
+        options=[
+            "sofa",
+            "gauge",
+            "bed",
+            "armchair",
+            "barrels",
+            "lamp with a shade",
+            "door",
+            "boxes",
+            "human",
+            "desk",
+            "plant",
+            "roll container",
+            "dresser",
+        ],
+        expected_answer=["gauge", "barrels"],
+    ),
+    MultipleChoiceImageTaskInput(
+        question="What is in the image?",
+        images_paths=[IMG_PATH + "fallen_human_2.png"],
+        options=[
+            "sofa",
+            "gauge",
+            "bed",
+            "armchair",
+            "barrels",
+            "lamp with a shade",
+            "door",
+            "boxes",
+            "human",
+            "desk",
+            "plant",
+            "roll container",
+            "dresser",
+        ],
+        expected_answer=["boxes", "human", "roll container"],
+    ),
+    MultipleChoiceImageTaskInput(
+        question="What is in the image?",
+        images_paths=[IMG_PATH + "image_1.jpg"],
+        options=[
+            "gauge",
+            "bed",
+            "barrels",
+            "lamp with a shade",
+            "door",
+            "boxes",
+            "human",
+            "desk",
+            "plant",
+            "roll container",
+        ],
+        expected_answer=["bed", "lamp with a shade", "door", "desk"],
+    ),
+    MultipleChoiceImageTaskInput(
+        question="What is in the image?",
+        images_paths=[IMG_PATH + "image_4.jpg"],
+        options=[
+            "sofa",
+            "gauge",
+            "bed",
+            "armchair",
+            "barrels",
+            "lamp with a shade",
+            "door",
+            "boxes",
+            "human",
+            "desk",
+            "plant",
+            "roll container",
+            "dresser",
+            "TV",
+        ],
+        expected_answer=["sofa", "plant", "TV", "dresser"],
+    ),
+    MultipleChoiceImageTaskInput(
+        question="What is in the image?",
+        images_paths=[IMG_PATH + "image_7.jpg"],
+        options=[
+            "sofa",
+            "gauge",
+            "armchair",
+            "barrels",
+            "door",
+            "boxes",
+            "human",
+            "desk",
+            "plant",
+            "roll container",
+        ],
+        expected_answer=["sofa", "armchair", "door", "desk", "plant"],
+    ),
 ]
 
 
@@ -109,4 +378,19 @@ def get_spatial_tasks() -> List[ImageReasoningTask[BaseModel]]:
         )
         for input_item in false_response_inputs
     ]
-    return cast(List[ImageReasoningTask[BaseModel]], true_tasks + false_tasks)
+    quantity_tasks = [
+        QuantityImageTask(
+            task_input=input_item,
+        )
+        for input_item in quantity_response_inputs
+    ]
+    multiple_choice_tasks = [
+        MultipleChoiceImageTask(
+            task_input=input_item,
+        )
+        for input_item in multiple_choice_response_inputs
+    ]
+    return cast(
+        List[ImageReasoningTask[BaseModel]],
+        true_tasks + false_tasks + quantity_tasks + multiple_choice_tasks,
+    )
