feat: Add CSGO-2

README.md

@@ -15,7 +15,7 @@ It uses a neural network to detect highlights in the video-game frames.\
 
 # Supported games
 
-Currently it supports **[Valorant](https://playvalorant.com/)**, **[Overwatch](https://playoverwatch.com/)**.
+Currently it supports **[Valorant](https://playvalorant.com/)**, **[Overwatch](https://playoverwatch.com/)** and **[CSGO2](https://www.counter-strike.net/cs2)**.
 
 # Usage
 
@@ -59,7 +59,63 @@ The following settings are adjustable:
   - second-before: Seconds of gameplay included before the highlight.
   - second-after: Seconds of gameplay included after the highlight.
   - second-between-kills: Transition time between highlights. If the time between two highlights is less than this value, the both highlights will be merged.
-- game: Chosen game (either "valorant" or "overwatch")
+- game: Chosen game (either "valorant", "overwatch" or "csgo2")
+
+### Recommended settings
+
+I recommend you to use the trials and errors method to find the best settings for your videos.\
+Here are some settings that I found to work well for me:
+
+#### Valorant
+
+```json
+{
+  "neural-network": {
+    "confidence": 0.8
+  },
+  "clip": {
+    "framerate": 8,
+    "second-before": 4,
+    "second-after": 0.5,
+    "second-between-kills": 3
+  },
+  "game": "valorant"
+}
+```
+
+#### Overwatch
+
+```json
+{
+  "neural-network": {
+    "confidence": 0.6
+  },
+  "clip": {
+    "framerate": 8,
+    "second-before": 4,
+    "second-after": 3,
+    "second-between-kills": 5
+  },
+  "game": "overwatch"
+}
+```
+
+#### CSGO2
+
+```json
+{
+  "neural-network": {
+    "confidence": 0.7
+  },
+  "clip": {
+    "framerate": 8,
+    "second-before": 4,
+    "second-after": 1,
+    "second-between-kills": 3
+  },
+  "game": "csgo2"
+}
+```
 
 ## Run
 
@@ -139,3 +195,7 @@ Now `pre-commit` will run on every `git commit`.
 
 - `cd crispy-frontend && yarn && yarn dev`
 - `cd crispy-backend && pip install -Ir requirements-dev.txt && python -m api`
+
+## Test
+
+- `cd crispy-api && pytest`

crispy-api/api/__init__.py

@@ -11,20 +11,24 @@
 from montydb import MontyClient, set_storage
 from pydantic.json import ENCODERS_BY_TYPE
 
-from api.config import DATABASE_PATH, DEBUG, GAME, MUSICS, VIDEOS
+from api.config import DATABASE_PATH, DEBUG, FRAMERATE, GAME, MUSICS, VIDEOS
 from api.tools.AI.network import NeuralNetwork
 from api.tools.enums import SupportedGames
 from api.tools.filters import apply_filters  # noqa
 from api.tools.setup import handle_highlights, handle_musics
 
 ENCODERS_BY_TYPE[ObjectId] = str
 
+neural_network = NeuralNetwork(GAME)
+
 if GAME == SupportedGames.OVERWATCH:
-    neural_network = NeuralNetwork([10000, 120, 15, 2])
     neural_network.load("./assets/overwatch.npy")
 elif GAME == SupportedGames.VALORANT:
-    neural_network = NeuralNetwork([4000, 120, 15, 2], 0.01)
     neural_network.load("./assets/valorant.npy")
+elif GAME == SupportedGames.CSGO2:
+    neural_network.load("./assets/csgo2.npy")
+else:
+    raise ValueError(f"game {GAME} not supported")
 
 
 logging.getLogger("PIL").setLevel(logging.ERROR)
@@ -62,7 +66,7 @@ def is_tool_installed(ffmpeg_tool: str) -> None:
 @app.on_event("startup")
 async def setup_crispy() -> None:
     await handle_musics(MUSICS)
-    await handle_highlights(VIDEOS, GAME, framerate=8)
+    await handle_highlights(VIDEOS, GAME, framerate=FRAMERATE)
 
 
 @app.exception_handler(HTTPException)

crispy-api/api/__main__.py

@@ -1,15 +1,35 @@
 import argparse
 import asyncio
+import os
+import sys
 
 import uvicorn
 
 from api import init_database
-from api.config import DEBUG, HOST, PORT
+from api.config import (
+    DATASET_CSV_PATH,
+    DATASET_CSV_TEST_PATH,
+    DEBUG,
+    HOST,
+    NETWORK_OUTPUTS_PATH,
+    PORT,
+)
+from api.tools.AI.trainer import Trainer, test, train
 from api.tools.dataset import create_dataset
 from api.tools.enums import SupportedGames
 
 _parser = argparse.ArgumentParser()
+# Dataset
 _parser.add_argument("--dataset", action="store_true")
+
+# Trainer
+_parser.add_argument("--train", help="Train the network", action="store_true")
+_parser.add_argument("--test", help="Test the network", action="store_true")
+_parser.add_argument("--epoch", help="Number of epochs", type=int, default=1000)
+_parser.add_argument("--load", help="Load a trained network", action="store_true")
+_parser.add_argument("--path", help="Path to the network", type=str)
+
+# Game
 _parser.add_argument(
     "--game", type=str, choices=[game.value for game in SupportedGames]
 )
@@ -26,11 +46,34 @@ async def generate_dataset(game: SupportedGames) -> None:
 
 
 if __name__ == "__main__":
-    if not _args.dataset:
+    if not _args.dataset and not _args.train and not _args.test:
         uvicorn.run("api:app", host=HOST, port=PORT, reload=DEBUG, proxy_headers=True)
     else:
         game = SupportedGames(_args.game)
-        if not game:
-            raise ValueError("Game not supported")
+        if _args.dataset:
+            if not game:
+                raise ValueError("Game not supported")
+
+            asyncio.run(generate_dataset(game))
+        else:
+            trainer = Trainer(game, 0.01)
+
+            if _args.load:
+                trainer.load(_args.path)
+            else:
+                trainer.initialize_weights()
+
+            print(trainer)
+            if _args.train:
+                if not os.path.exists(NETWORK_OUTPUTS_PATH):
+                    os.makedirs(NETWORK_OUTPUTS_PATH)
+                train(
+                    _args.epoch, trainer, DATASET_CSV_PATH, True, NETWORK_OUTPUTS_PATH
+                )
+
+            if _args.test:
+                if not _args.load and not _args.train:
+                    print("You need to load a trained network")
+                    sys.exit(1)
 
-        asyncio.run(generate_dataset(game))
+                sys.exit(not test(trainer, DATASET_CSV_TEST_PATH))

crispy-api/api/config.py

@@ -14,7 +14,8 @@
 
 ASSETS = "assets"
 SILENCE_PATH = os.path.join(ASSETS, "silence.mp3")
-DOT_PATH = os.path.join(ASSETS, "dot.png")
+VALORANT_MASK_PATH = os.path.join(ASSETS, "valorant-mask.png")
+CSGO2_MASK_PATH = os.path.join(ASSETS, "csgo2-mask.png")
 
 BACKUP = "backup"
 
@@ -24,7 +25,10 @@
 MUSICS = os.path.join(RESOURCES, "musics")
 
 DATASET_PATH = "dataset"
-DATASET_VALUES_PATH = "dataset-values.json"
+DATASET_VALUES_PATH = os.path.join(DATASET_PATH, "dataset-values.json")
+DATASET_CSV_PATH = os.path.join(DATASET_PATH, "result.csv")
+DATASET_CSV_TEST_PATH = os.path.join(DATASET_PATH, "test.csv")
+NETWORK_OUTPUTS_PATH = "outputs"
 
 DATABASE_PATH = ".data"
 

crispy-api/api/models/highlight.py

@@ -7,14 +7,16 @@
 from mongo_thingy import Thingy
 from PIL import Image, ImageFilter, ImageOps
 
-from api.config import BACKUP, DOT_PATH
+from api.config import BACKUP, CSGO2_MASK_PATH, VALORANT_MASK_PATH
 from api.models.filter import Filter
 from api.models.segment import Segment
 from api.tools.audio import silence_if_no_audio
 from api.tools.enums import SupportedGames
 from api.tools.ffmpeg import merge_videos
 
 logger = logging.getLogger("uvicorn")
+valorant_mask = Image.open(VALORANT_MASK_PATH)
+csgo2_mask = Image.open(CSGO2_MASK_PATH)
 
 
 class Highlight(Thingy):
@@ -130,9 +132,7 @@ def _apply_filter_and_do_operations(
 
             image = image.crop((1, 1, image.width - 2, image.height - 2))
 
-            dot = Image.open(DOT_PATH)
-
-            image.paste(dot, (0, 0), dot)
+            image.paste(valorant_mask, (0, 0), valorant_mask)
 
             left = image.crop((0, 0, 25, 60))
             right = image.crop((95, 0, 120, 60))
@@ -162,13 +162,31 @@ def post_process(image: Image) -> Image:
             post_process, (899, 801, 122, 62), framerate=framerate
         )
 
+    async def extract_csgo2_images(self, framerate: int = 4) -> bool:
+        def post_process(image: Image) -> Image:
+            image = ImageOps.grayscale(
+                image.filter(ImageFilter.FIND_EDGES).filter(
+                    ImageFilter.EDGE_ENHANCE_MORE
+                )
+            )
+            final = Image.new("RGB", (100, 100))
+            final.paste(image, (0, 0))
+            final.paste(csgo2_mask, (0, 0), csgo2_mask)
+            return final
+
+        return await self.extract_images(
+            post_process, (930, 925, 100, 100), framerate=framerate
+        )
+
     async def extract_images_from_game(
         self, game: SupportedGames, framerate: int = 4
     ) -> bool:
         if game == SupportedGames.OVERWATCH:
             return await self.extract_overwatch_images(framerate)
         elif game == SupportedGames.VALORANT:
             return await self.extract_valorant_images(framerate)
+        elif game == SupportedGames.CSGO2:
+            return await self.extract_csgo2_images(framerate)
         else:
             raise NotImplementedError
 

crispy-api/api/tools/AI/network.py

@@ -3,14 +3,22 @@
 import numpy as np
 import scipy.special
 
+from api.tools.enums import SupportedGames
+
+NetworkResolution = {
+    SupportedGames.VALORANT: [4000, 120, 15, 2],
+    SupportedGames.OVERWATCH: [10000, 120, 15, 2],
+    SupportedGames.CSGO2: [10000, 120, 15, 2],
+}
+
 
 class NeuralNetwork:
     """
     Neural network to predict if a kill is on the image
     """
 
-    def __init__(self, nodes: List[int], learning_rate: float = 0.01) -> None:
-        self.nodes = nodes
+    def __init__(self, game: SupportedGames, learning_rate: float = 0.01) -> None:
+        self.nodes = NetworkResolution[game]
         self.learning_rate = learning_rate
         self.weights: List[Any] = []
         self.activation_function = lambda x: scipy.special.expit(x)
@@ -55,7 +63,7 @@ def _train(self, inputs: List[float], targets: Any) -> Tuple[int, int, int]:
         for i in range(len(self.nodes) - 1 - 1, 0, -1):
             errors.insert(0, np.dot(self.weights[i].T, errors[0]))
 
-        # ten times more likely to be not be kill
+        # five times more likely to be not be kill
         # so we mitigate the error
         if expected == 0:
             errors = [e / 5 for e in errors]