pollen-robotics/pollen-vision

Different format of Open AI object descriptions generated.

Opened this issue · 0 comments

When I run the object description generator locally I get the usual format of objects descriptions. For 'croissant' for example:

"croissant": [
    "A croissant is a crescent-shaped pastry with a flaky and golden-brown crust.",
    "The identifying characteristic of a croissant is its crescent shape and flaky texture.",
    "It looks like a crescent-shaped pastry with a light and airy interior.",
}

But when I run it on Colab, I get a different format:

{
  "croissant": [
    {
      "description": "It is a crescent-shaped pastry with a flaky texture and golden-brown color.",
      "identifying_characteristics": "Crescent shape, flaky texture, golden-brown color",
      "visual_characteristics": "Crescent shape, flakiness, golden-brown exterior"
    },
    {
      "description": "A buttery, crescent-shaped bread roll with a crusty exterior and soft interior.",
      "identifying_characteristics": "Buttery, crescent shape, crusty exterior, soft interior",
      "visual_characteristics": "Crusty exterior, soft interior, crescent shape"
    },

Because the format is different, I cannot instanciate a RAM wrapper with this descriptions. I used the same OPENAI API key in both cases.