library(kuzco)
test_img <- file.path(system.file(package = "kuzco"), "img/test_img.jpg") multilingual support
Some models are capable of returning outputs in multiple languages. As a result, it became necessary to include a language specification to ensure certain models responded in English. Once this step was in place, switching the output language for specific computer vision tasks became straightforward & easy to implement.
English Results
llm_image_classification(
llm_model = "qwen2.5vl",
image = test_img,
backend = 'ellmer',
provider = 'ollama',
language = 'English'
)
#> image_classification primary_object secondary_object
#> 1 Portrait of a puppy puppy
#> image_description
#> 1 A close-up portrait of a puppy with black and white fur is the main subject, with a red plaid blanket in the background.
#> image_colors image_proba_names
#> 1 ["#000000", "#FFFFFF", "#FF0000", "#0000FF", "#A52A2A"]
#> image_proba_values
#> 1 [0.6, 0.15, 0.08, 0.07, 0.02]Spanish Results
llm_image_classification(
llm_model = "qwen2.5vl",
image = test_img,
backend = 'ellmer',
provider = 'ollama',
language = 'Spanish'
)
#> image_classification primary_object secondary_object
#> 1 perrito perrito pañal
#> image_description
#> 1 Un adorable perrito de color negro y blanco está echado en la cama. El perrito está mirando directamente a la cámara. Se ven texturas sutiles en el pelaje y un fondo de colcha de cuadros rojos.
#> image_colors image_proba_names
#> 1 [#000000, #FFFFFF, #D60000, #FF0000, #0000FF] perrito
#> image_proba_values
#> 1 :[0.8,0.0,0.0,0.0,0.0]German Results
llm_image_classification(
llm_model = "qwen2.5vl",
image = test_img,
backend = 'ellmer',
provider = 'ollama',
language = 'German'
)
#> image_classification primary_object secondary_object
#> 1 Kinderbetten Hund
#> image_description
#> 1 Der Bildinhalt zeigt einen Hund, der in der Nähe eines Kinderbetts liegt.
#> image_colors
#> 1 [⌘A87, ⌘AADE, ⌘545417, ⌘545417, ⌘AADE, ⌘A87, ⌘38182A, ⌘A87]
#> image_proba_names
#> 1 Hund, Bett, Tier, Kind, Hundebett, Tiergesicht, Fell, Haar, Tieraugen
#> image_proba_values
#> 1 [0.42, 0.106, 0.08, 0.075, 0.045, 0.035, 0.034, 0.09, 0.07]Chinese Results
llm_image_classification(
llm_model = "qwen2.5vl",
image = test_img,
backend = 'ellmer',
provider = 'ollama',
language = 'Mandarin Chinese'
)
#> image_classification primary_object secondary_object
#> 1 pet portrait puppy
#> image_description
#> 1 这是一张黑白相间的小狗的特写照片，它有着独特的眼睛和湿润的鼻子。
#> image_colors image_proba_names
#> 1 ["#990000", "#0000FF", "#FFC0CB", "#000000", "#FFFFFF"] [
#> image_proba_values
#> 1 [