Skip to main content

semantic search

semantic search allows you to find photos using natural language descriptions instead of relying on manual tags or filenames.

how it works​

silo uses clip (contrastive language-image pre-training) to understand the content of your images:

  1. clip generates embeddings for each image
  2. your search query is converted to an embedding
  3. faiss finds the most similar images
  4. results are ranked by relevance

search types​

describe what you see in the image:

  • "sunset over mountains"
  • "dogs playing in snow"
  • "person wearing red shirt"
  • "trees in autumn"
  • "camper in the forest"
  • "husky dog"

search example: trees

search example: camper

search example: husky

search for actions and activities:

  • "people dancing"
  • "kids playing soccer"
  • "cooking in kitchen"
  • "swimming in pool"

find photos by location type:

  • "beach photos"
  • "mountain hiking"
  • "city skyline"
  • "indoor restaurant"

search for specific objects:

  • "laptop on desk"
  • "coffee mug"
  • "bicycle"
  • "flowers in vase"

once you've named people and animal clusters, you can search for them specifically:

naming clusters​

  1. go to the "people" or "animals" tab
  2. click on a cluster card
  3. enter a name (e.g., "blake", "spot")
  4. click save

searching named clusters​

now you can search using those names:

  • "blake and spot on the beach"
  • "spot playing in the yard"
  • "blake at birthday party"

combine names with activities, locations, and objects for powerful searches:

  • "blake and grandma baking cookies"
  • "spot running in the park"
  • "family photo with everyone"

search tips​

be specific​

more specific queries yield better results:

  • ❌ "outside"
  • ✅ "mountain landscape with snow"

use descriptive words​

include visual details:

  • ❌ "party"
  • ✅ "birthday party with cake and balloons"

combine concepts​

mix multiple search terms:

  • "black cat on wooden table"
  • "sunset over ocean with palm trees"
  • "group of people hiking in forest"

try variations​

if you don't find what you're looking for:

  • "dog" → "puppy", "canine", "pet"
  • "car" → "vehicle", "automobile"
  • "beach" → "ocean", "seaside", "coast"

search results​

results are displayed as a grid of thumbnails:

  • click any image to view full size
  • scroll to load more results
  • results are sorted by relevance score

limitations​

semantic search works best for:

  • clear, well-lit images
  • recognizable objects and scenes
  • common activities and locations

it may struggle with:

  • abstract art
  • very dark or blurry images
  • highly specific technical content
  • images with minimal visual information

text search​

silo also searches text extracted via ocr:

  • search for text visible in images
  • find screenshots with specific content
  • locate documents by visible text

example searches:

  • "text containing invoice"
  • "sign that says coffee shop"
  • "handwritten note"