whichframe / README.md
David Chuan-En Lin
init 7415fc4
1
---
2
title: Which Frame?
3
emoji: ๐Ÿ”
4
colorFrom: purple
5
colorTo: purple
6
sdk: streamlit
7
app_file: whichframe.py
8
---
9
10
# Which Frame?
11
12
Search a video **semantically** with AI. For example, try a natural language search query like "a person with sunglasses". You can also search with images like Google's reverse image search and also a combined text + image. The underlying querying is powered by OpenAIโ€™s CLIP neural network for "zero-shot" image classification.
13
14
---
15
16
## Try it out!
17
18
http://whichframe.chuanenlin.com
19
20
---
21
22
## Setting up
23
24
1.  Clone the repository.
25
26
```python
27
git clone https://github.com/chuanenlin/whichframe.git
28
cd whichframe
29
```
30
31
2.  Install package dependencies.
32
33
```python
34
pip install -r requirements.txt
35
```
36
37
3.  Run the app.
38
39
```python
40
streamlit run whichframe.py
41
```
42
43
---
44
45
## Examples
46
47
### ๐Ÿ”ค Text Search
48
49
#### Query
50
51
"three red cars side by side"
52
53
#### Result
54
55
![three-red-cars-side-by-side](examples/three-red-cars-side-by-side.jpeg)
56
57
### ๐ŸŒ… Image Search
58
59
#### Query
60
61
![police-car-query](examples/helicopter-query.jpeg)
62
63
#### Result
64
65
![police-car-result](examples/helicopter-result.jpeg)
66
67
### ๐Ÿ”ค Text + ๐ŸŒ… Image Search
68
69
#### Query
70
71
"a red subaru" +
72
73
![police-car-query](examples/police-car-query.jpeg)
74
75
#### Result
76
77
![subaru-and-police-car-result](examples/subaru-and-police-car-result.jpeg)
78