Add files using upload-large-folder tool

.gitattributes

@@ -33,3 +33,25 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+geneval/00118/samples/00002.png filter=lfs diff=lfs merge=lfs -text
+geneval/00118/samples/00001.png filter=lfs diff=lfs merge=lfs -text
+geneval/00118/samples/00000.png filter=lfs diff=lfs merge=lfs -text
+geneval/00062/samples/00003.png filter=lfs diff=lfs merge=lfs -text
+geneval/00118/samples/00003.png filter=lfs diff=lfs merge=lfs -text
+geneval/00062/samples/00002.png filter=lfs diff=lfs merge=lfs -text
+geneval/00062/samples/00000.png filter=lfs diff=lfs merge=lfs -text
+geneval/00062/samples/00001.png filter=lfs diff=lfs merge=lfs -text
+geneval/00181/samples/00003.png filter=lfs diff=lfs merge=lfs -text
+geneval/00118/grid.png filter=lfs diff=lfs merge=lfs -text
+geneval/00181/samples/00001.png filter=lfs diff=lfs merge=lfs -text
+tokenizer_config.json filter=lfs diff=lfs merge=lfs -text
+geneval/00181/samples/00000.png filter=lfs diff=lfs merge=lfs -text
+geneval/00181/samples/00002.png filter=lfs diff=lfs merge=lfs -text
+geneval/00181/grid.png filter=lfs diff=lfs merge=lfs -text
+geneval/00015/samples/00000.png filter=lfs diff=lfs merge=lfs -text
+geneval/00062/grid.png filter=lfs diff=lfs merge=lfs -text
+geneval/00015/samples/00001.png filter=lfs diff=lfs merge=lfs -text
+geneval/00015/samples/00003.png filter=lfs diff=lfs merge=lfs -text
+geneval/00015/samples/00002.png filter=lfs diff=lfs merge=lfs -text
+geneval/00528/samples/00003.png filter=lfs diff=lfs merge=lfs -text
+geneval/00528/samples/00000.png filter=lfs diff=lfs merge=lfs -text

geneval/00006/metadata.jsonl

@@ -0,0 +1 @@
+{"tag": "single_object", "include": [{"class": "fork", "count": 1}], "prompt": "a photo of a fork", "detailed_caption": "A clear and simple photo of a single fork placed on a flat, neutral-colored surface. The fork is made of stainless steel, with four evenly spaced tines and a smooth, polished handle that reflects light subtly. The background is plain, ensuring that the focus remains solely on the fork.", "index": "00006"}

geneval/00034/metadata.jsonl

@@ -0,0 +1 @@
+{"tag": "single_object", "include": [{"class": "cell phone", "count": 1}], "prompt": "a photo of a cell phone", "detailed_caption": "A clear photo of a modern cell phone placed on a flat surface. The cell phone has a sleek design with a large touchscreen display that reflects light subtly. The back of the phone shows a minimalist finish, with a visible camera lens in the corner. The background is neutral and simple, ensuring the focus remains solely on the cell phone.", "index": "00034"}

geneval/00043/metadata.jsonl

@@ -0,0 +1 @@
+{"tag": "single_object", "include": [{"class": "scissors", "count": 1}], "prompt": "a photo of a scissors", "detailed_caption": "A clear photo of a pair of scissors placed on a plain surface. The scissors feature shiny metal blades with straight edges and have simple, comfortable handles. The background is minimal and unobtrusive, ensuring the focus remains solely on the scissors.", "index": "00043"}

geneval/00049/metadata.jsonl

@@ -0,0 +1 @@
+{"tag": "single_object", "include": [{"class": "person", "count": 1}], "prompt": "a photo of a person", "detailed_caption": "A clear photo of a person standing outdoors in a natural setting. The individual is dressed in casual clothing, with a friendly expression and natural pose. Behind them, the background features soft focus elements of greenery and blue sky, creating a peaceful and welcoming atmosphere that keeps the attention on the person.", "index": "00049"}

geneval/00071/metadata.jsonl

@@ -0,0 +1 @@
+{"tag": "single_object", "include": [{"class": "bus", "count": 1}], "prompt": "a photo of a bus", "detailed_caption": "A clear photo of a large, modern bus parked on a city street. The bus is painted in vibrant colors with large windows and visible headlights. Its long body shows multiple seats inside through the windows, and the city's urban environment can be seen subtly in the background. The road beneath the bus is smooth, and the sky is clear, making the bus the central focus of the image.", "index": "00071"}

geneval/00095/metadata.jsonl

@@ -0,0 +1 @@
+{"tag": "two_object", "include": [{"class": "book", "count": 1}, {"class": "laptop", "count": 1}], "prompt": "a photo of a book and a laptop", "detailed_caption": "A clear photo of a book and a laptop placed side by side on a clean desk. The book is open, revealing printed text on its pages, while the laptop is closed, showcasing its sleek exterior. The desk surface is uncluttered, and the background is simple, ensuring the focus is on the book and the laptop.", "index": "00095"}

geneval/00101/metadata.jsonl

@@ -0,0 +1 @@
+{"tag": "two_object", "include": [{"class": "potted plant", "count": 1}, {"class": "backpack", "count": 1}], "prompt": "a photo of a potted plant and a backpack", "detailed_caption": "A clear photo of a potted plant and a backpack placed side by side on a flat surface. The potted plant, with lush green leaves, sits in a simple, round pot, while the backpack is sturdy and features multiple compartments with visible zippers. The background is plain, keeping the attention on the potted plant and the backpack.", "index": "00101"}

geneval/00120/metadata.jsonl

@@ -0,0 +1 @@
+{"tag": "two_object", "include": [{"class": "parking meter", "count": 1}, {"class": "teddy bear", "count": 1}], "prompt": "a photo of a parking meter and a teddy bear", "detailed_caption": "A clear photo of a parking meter and a teddy bear positioned side by side on a simple surface. The parking meter is a standard model with a metallic body and a visible display, while the teddy bear is plush and soft, featuring a classic design with brown fur and stitched eyes. The background is plain, keeping the focus on the parking meter and the teddy bear.", "index": "00120"}

geneval/00133/metadata.jsonl

@@ -0,0 +1 @@
+{"tag": "two_object", "include": [{"class": "computer keyboard", "count": 1}, {"class": "microwave", "count": 1}], "prompt": "a photo of a computer keyboard and a microwave", "detailed_caption": "A clear photo of a computer keyboard and a microwave placed on a flat surface. The computer keyboard is standard-sized with a black finish and clearly visible keys, while the microwave is compact and features a metallic exterior with a digital display and buttons. The background is simple and unobtrusive, ensuring full attention is on the computer keyboard and the microwave.", "index": "00133"}

geneval/00139/metadata.jsonl

@@ -0,0 +1 @@
+{"tag": "two_object", "include": [{"class": "toothbrush", "count": 1}, {"class": "bench", "count": 1}], "prompt": "a photo of a toothbrush and a bench", "detailed_caption": "A clear photo of a toothbrush and a bench positioned side by side in a simple setting. The toothbrush features a blue and white plastic handle with bristles that are slightly angled. The bench is wooden with a natural finish, showing its sturdy legs and slatted seat. The background is neutral, ensuring the focus remains on the toothbrush and the bench.", "index": "00139"}

geneval/00144/metadata.jsonl

@@ -0,0 +1 @@
+{"tag": "two_object", "include": [{"class": "sports ball", "count": 1}, {"class": "cow", "count": 1}], "prompt": "a photo of a sports ball and a cow", "detailed_caption": "A clear photo of a sports ball and a cow standing on a grassy field. The sports ball, featuring distinct patterns and colors typically associated with team sports, is positioned near the cow. The cow stands calmly, showcasing its natural markings and features. The background is simple, with the green grass providing a natural setting, allowing the focus to remain on the sports ball and the cow.", "index": "00144"}

geneval/00157/metadata.jsonl

@@ -0,0 +1 @@
+{"tag": "two_object", "include": [{"class": "stop sign", "count": 1}, {"class": "fork", "count": 1}], "prompt": "a photo of a stop sign and a fork", "detailed_caption": "A clear photo featuring a stop sign and a fork placed against a plain background. The stop sign, with its distinctive red octagonal shape and bold white lettering, is positioned prominently in the frame. Next to it, a simple metal fork with four tines is placed horizontally. The background is neutral, allowing the stop sign's vivid color and the fork's metallic sheen to stand out as the primary focus of the image.", "index": "00157"}

geneval/00176/metadata.jsonl

@@ -0,0 +1 @@
+{"tag": "two_object", "include": [{"class": "person", "count": 1}, {"class": "traffic light", "count": 1}], "prompt": "a photo of a person and a traffic light", "detailed_caption": "A clear photo of a person standing next to a traffic light on a city street. The person is casually dressed and looking at the light, which features the classic red, yellow, and green signals. The setting is an urban environment, but the background is kept simple and unobtrusive to maintain focus on the person and the traffic light.", "index": "00176"}

geneval/00192/metadata.jsonl

@@ -0,0 +1 @@
+{"tag": "counting", "include": [{"class": "cup", "count": 3}], "exclude": [{"class": "cup", "count": 4}], "prompt": "a photo of three cups", "detailed_caption": "A clear photo of three cups arranged neatly in a row on a smooth surface. Each cup is identical in shape but features a different color: one is blue, another is green, and the third is yellow. The cups have a simple, classic design with smooth exteriors. The background is plain and unobtrusive, ensuring the focus stays on the three colorful cups.", "index": "00192"}

geneval/00198/metadata.jsonl

@@ -0,0 +1 @@
+{"tag": "counting", "include": [{"class": "train", "count": 2}], "exclude": [{"class": "train", "count": 3}], "prompt": "a photo of two trains", "detailed_caption": "A clear photo of two trains positioned side by side on parallel tracks. Each train features a distinct design and color scheme, showcasing their unique styles and details. The scene is set in an open area with tracks extending into the distance, and a simple background ensures that the focus is maintained on the two trains.", "index": "00198"}

geneval/00202/metadata.jsonl

@@ -0,0 +1 @@
+{"tag": "counting", "include": [{"class": "snowboard", "count": 3}], "exclude": [{"class": "snowboard", "count": 4}], "prompt": "a photo of three snowboards", "detailed_caption": "A clear photo of three snowboards standing upright next to each other against a snowy backdrop. Each snowboard features distinct designs and vibrant colors, showcasing a variety of patterns and styles. The snow-covered ground and soft, overcast sky create a chilly winter atmosphere, with the focus remaining on the three uniquely designed snowboards.", "index": "00202"}

geneval/00208/metadata.jsonl

@@ -0,0 +1 @@
+{"tag": "counting", "include": [{"class": "zebra", "count": 3}], "exclude": [{"class": "zebra", "count": 4}], "prompt": "a photo of three zebras", "detailed_caption": "A vivid photo of three zebras standing together on a grassy savanna. The zebras' distinctive black and white striped patterns create a striking visual as they stand close to one another. The grass beneath them is short and dry, typical of a savanna environment. The sky above is clear and blue, providing a simple, natural backdrop that keeps the focus on the three zebras.", "index": "00208"}

geneval/00223/metadata.jsonl

@@ -0,0 +1 @@
+{"tag": "counting", "include": [{"class": "stop sign", "count": 4}], "exclude": [{"class": "stop sign", "count": 5}], "prompt": "a photo of four stop signs", "detailed_caption": "A clear photo of four stop signs standing in a row, each featuring the familiar octagonal shape with bright red color and bold white letters spelling \"STOP.\" The signs are mounted on posts at uniform heights, set against a neutral background that highlights their vibrant color and clear message. The image focuses on the repetition and alignment of the four stop signs.", "index": "00223"}

geneval/00229/metadata.jsonl

@@ -0,0 +1 @@
+{"tag": "counting", "include": [{"class": "wine glass", "count": 3}], "exclude": [{"class": "wine glass", "count": 4}], "prompt": "a photo of three wine glasses", "detailed_caption": "A clear photo of three wine glasses arranged in a row on a flat surface. Each glass is elegant, with slender stems and rounded bowls, reflecting light subtly. The surface beneath them is plain, and the background is neutral, allowing the focus to remain on the three wine glasses and their graceful arrangement.", "index": "00229"}

geneval/00230/metadata.jsonl

@@ -0,0 +1 @@
+{"tag": "counting", "include": [{"class": "broccoli", "count": 4}], "exclude": [{"class": "broccoli", "count": 5}], "prompt": "a photo of four broccolis", "detailed_caption": "A clear photo of four broccoli heads neatly arranged on a simple, flat surface. Each broccoli has a vibrant green color with tightly clustered florets and sturdy stalks. The background is plain and unadorned, keeping the focus on the four broccoli heads and their fresh, natural appearance.", "index": "00230"}

geneval/00237/metadata.jsonl

@@ -0,0 +1 @@
+{"tag": "counting", "include": [{"class": "car", "count": 2}], "exclude": [{"class": "car", "count": 3}], "prompt": "a photo of two cars", "detailed_caption": "A clear photo of two cars parked side by side on a paved surface. One car is a sleek blue sedan, and the other is a compact white SUV, both exhibiting modern designs and shiny exteriors. The background is simple and unobtrusive, ensuring that the focus remains on the two cars and their distinct styles.", "index": "00237"}

geneval/00240/metadata.jsonl

@@ -0,0 +1 @@
+{"tag": "counting", "include": [{"class": "pizza", "count": 3}], "exclude": [{"class": "pizza", "count": 4}], "prompt": "a photo of three pizzas", "detailed_caption": "A photo featuring three pizzas arranged side by side on a flat, simple surface. Each pizza has a unique combination of toppings, with melted cheese, pepperoni slices, vibrant vegetables, and fresh herbs. The crusts appear golden and perfectly baked. The background is plain and unobtrusive, keeping the focus on the delicious trio of pizzas.", "index": "00240"}

geneval/00247/metadata.jsonl

@@ -0,0 +1 @@
+{"tag": "counting", "include": [{"class": "bird", "count": 3}], "exclude": [{"class": "bird", "count": 4}], "prompt": "a photo of three birds", "detailed_caption": "A clear photo of three birds perched closely together on a thin branch. Each bird has distinct plumage, showcasing a variety of colors and patterns, from vibrant to more muted tones. The background is soft-focused, with hints of greenery and sky, emphasizing the birds as the central subject of the image.", "index": "00247"}

geneval/00254/metadata.jsonl

@@ -0,0 +1 @@
+{"tag": "counting", "include": [{"class": "bench", "count": 4}], "exclude": [{"class": "bench", "count": 5}], "prompt": "a photo of four benchs", "detailed_caption": "A clear photo of four benches positioned in a row. Each bench has a simple, classic design with wooden slats and metal legs, placed on a paved surface. The setting is minimalistic, with a plain background, directing attention to the uniform arrangement of the four benches.", "index": "00254"}

geneval/00275/metadata.jsonl

@@ -0,0 +1 @@
+{"tag": "colors", "include": [{"class": "tv remote", "count": 1, "color": "yellow"}], "prompt": "a photo of a yellow tv remote", "detailed_caption": "A clear photo of a yellow TV remote placed on a flat surface. The remote has a sleek and modern design with a vibrant yellow color, featuring an array of buttons for various functions. The surface and background are plain and uncluttered, keeping the focus entirely on the yellow TV remote.", "index": "00275"}

geneval/00291/metadata.jsonl

@@ -0,0 +1 @@
+{"tag": "colors", "include": [{"class": "airplane", "count": 1, "color": "yellow"}], "prompt": "a photo of a yellow airplane", "detailed_caption": "A clear photo of a yellow airplane captured in flight against a bright blue sky. The airplane has a sleek, aerodynamic design with distinct wings and tail, painted in a vibrant yellow color that stands out against the backdrop. The sky is clear, offering a perfect contrast that highlights the airplane's features and color. There are no additional objects in the frame, keeping the focus solely on the yellow airplane as it soars through the sky.", "index": "00291"}

geneval/00305/metadata.jsonl

@@ -0,0 +1 @@
+{"tag": "colors", "include": [{"class": "bear", "count": 1, "color": "brown"}], "prompt": "a photo of a brown bear", "detailed_caption": "A detailed photo of a brown bear standing in a natural setting. The bear is captured in a dynamic pose, showcasing its thick fur and sturdy build. The background is composed of a simple natural landscape, suggesting a forest or a clearing, with minimal distractions to keep the focus on the brown bear.", "index": "00305"}

geneval/00324/metadata.jsonl

@@ -0,0 +1 @@
+{"tag": "colors", "include": [{"class": "teddy bear", "count": 1, "color": "black"}], "prompt": "a photo of a black teddy bear", "detailed_caption": "A clear photo of a plush black teddy bear sitting upright on a plain surface. The teddy bear features soft, fuzzy fur with stitched facial features and a slightly rounded body, giving it a cute and cuddly appearance. Its backdrop is plain and neutral, ensuring the attention stays focused solely on the black teddy bear.", "index": "00324"}

geneval/00330/metadata.jsonl

@@ -0,0 +1 @@
+{"tag": "colors", "include": [{"class": "dining table", "count": 1, "color": "blue"}], "prompt": "a photo of a blue dining table", "detailed_caption": "A clear photo of a blue dining table set in a simple room. The table has a smooth surface with clean lines and a rich blue color that stands out. The room is minimally furnished, ensuring the focus remains on the dining table without any distractions in the background.", "index": "00330"}

geneval/00337/metadata.jsonl

@@ -0,0 +1 @@
+{"tag": "colors", "include": [{"class": "carrot", "count": 1, "color": "blue"}], "prompt": "a photo of a blue carrot", "detailed_caption": "A clear photo of a carrot creatively colored blue, resting on a plain white surface. The carrot retains its natural shape and texture, with the vibrant blue hue adding an imaginative twist. The background is kept simple and unadorned, ensuring that the striking blue carrot remains the focal point of the image.", "index": "00337"}