diff --git a/src/conversion_scripts/convert_marc.py b/src/conversion_scripts/convert_marc.py
index e4e037460c3267963747f3b1d40dd07eef813cf2..5640ca6775ee46f46f4a18a37d69ec16a52b98f4 100644
--- a/src/conversion_scripts/convert_marc.py
+++ b/src/conversion_scripts/convert_marc.py
@@ -19,7 +19,7 @@ import robofish.io
def convertTrajectory(path, save_path, categories):
ar = pd.read_csv(path, sep=";").to_numpy()
- new = robofish.io.File(world_size_cm=[100, 100], frequency_hz=20)
+ new = robofish.io.File(world_size_cm=[100, 100], frequency_hz=25)
# convert x,y from m to cm
ar[:, [0, 1, 3, 4]] = ar[:, [0, 1, 3, 4]] * 100
diff --git a/src/robofish/evaluate/evaluate.py b/src/robofish/evaluate/evaluate.py
index 6670645a03ce8a6460a6e4febbbc63b3c8e32e74..a5719a5037c76b7ce76256d9481ea5ad1e342824 100644
--- a/src/robofish/evaluate/evaluate.py
+++ b/src/robofish/evaluate/evaluate.py
@@ -87,7 +87,7 @@ def evaluate_speed(
if labels is None:
labels = paths
- plt.hist(speeds, bins=20, label=labels, density=True, range=[0, 50])
+ plt.hist(speeds, bins=20, label=labels, density=True, range=[0, 25])
plt.title("Agent speeds")
plt.xlabel("Speed [cm/s]")
plt.ylabel("Frequency")
@@ -147,7 +147,7 @@ def evaluate_turn(
labels = paths
# TODO: Quantil range
- plt.hist(turns, bins=40, label=labels, density=True, range=[-30, 30])
+ plt.hist(turns, bins=41, label=labels, density=True, range=[-30, 30])
plt.title("Agent turns")
plt.xlabel("Change in orientation [Degree / timestep at %dhz]" % frequency)
plt.ylabel("Frequency")
@@ -193,8 +193,8 @@ def evaluate_orientation(
world_size[0] / 2,
world_size[1] / 2,
]
- xbins = np.linspace(world_bounds[0], world_bounds[2], 10)
- ybins = np.linspace(world_bounds[1], world_bounds[3], 10)
+ xbins = np.linspace(world_bounds[0], world_bounds[2], 11)
+ ybins = np.linspace(world_bounds[1], world_bounds[3], 11)
ret_1 = stats.binned_statistic_2d(
poses[:, 0], poses[:, 1], poses[:, 2], "mean", bins=[xbins, ybins]
)
@@ -226,13 +226,13 @@ def evaluate_orientation(
plot = ax[i].pcolormesh(
xx,
- yy * (-1),
+ yy,
np.arctan2(s_2, s_1).T,
vmin=-np.pi,
vmax=np.pi,
cmap="twilight",
)
- # cbar = plt.colorbar(plot, ax=ax[i], pad=0.015, aspect=10)
+ cbar = plt.colorbar(plot, ax=ax[i], pad=0.015, aspect=10)
show_values(plot)
if save_path is None:
@@ -419,7 +419,10 @@ def evaluate_tankpositions(
ax[i].set_xlim(-world_bounds[i][0] / 2, world_bounds[i][0] / 2)
ax[i].set_ylim(-world_bounds[i][1] / 2, world_bounds[i][1] / 2)
- sns.kdeplot(x=x_pos[i], y=y_pos[i] * (-1), n_levels=25, shade=True, ax=ax[i])
+ ax[i].set_xlabel("x [cm]")
+ ax[i].set_ylabel("y [cm]")
+
+ sns.kdeplot(x=x_pos[i], y=y_pos[i], n_levels=25, shade=True, ax=ax[i])
if save_path is None:
plt.show()
@@ -450,21 +453,26 @@ def evaluate_trajectories(
pos = []
world_bounds = []
for k, files in enumerate(files_per_path):
+ path_poses = []
for p, file in files.items():
poses = file.select_entity_poses(
None if predicate is None else predicate[k]
)
world_bounds.append(file.attrs["world_size_cm"])
- path_pos = {
- fish: pd.DataFrame({"x": poses[fish, :, 0], "y": poses[fish, :, 1]})
- for fish in range(len(poses))
- }
- combined = pd.concat(
- [
- path_pos[fish].assign(Agent=f"Agent {fish}")
- for fish in path_pos.keys()
- ]
- )
+ path_poses.append(poses[:, :, :2])
+ poses = np.concatenate(path_poses, axis = 1)
+
+ path_pos = {
+ fish: pd.DataFrame({"x": poses[fish, :, 0], "y": poses[fish, :, 1]})
+ for fish in range(len(poses))
+ }
+ combined = pd.concat(
+ [
+ path_pos[fish].assign(Agent=f"Agent {fish}")
+ for fish in path_pos.keys()
+ ]
+ )
+
pos.append((path_pos, combined))
fig, ax = plt.subplots(1, len(pos), figsize=(len(pos) * 8, 8))
@@ -481,7 +489,7 @@ def evaluate_trajectories(
ax[i].set_title("Trajectories (%s)" % labels[i])
ax[i].set_xlim(-world_bounds[i][0] / 2, world_bounds[i][0] / 2)
ax[i].set_ylim(-world_bounds[i][1] / 2, world_bounds[i][1] / 2)
- ax[i].invert_yaxis()
+ # ax[i].invert_yaxis()
ax[i].xaxis.set_ticks_position("top")
ax[i].xaxis.set_label_position("top")
ax[i].yaxis.set_ticks_position("left")
@@ -557,11 +565,10 @@ def evaluate_positionVec(
worldBoundsX, worldBoundsY = max(worldBoundsX), max(worldBoundsY)
for i in range(len(posVec)):
- df = pd.DataFrame({"x": posVec[i][:, 0], "y": posVec[i][:, 1] * (-1)})
+ df = pd.DataFrame({"x": posVec[i][:, 0], "y": posVec[i][:, 1]})
grid = sns.displot(df, x="x", y="y", binwidth=(10, 10), cbar=True)
grid.axes[0, 0].set_xlabel("x [cm]")
grid.axes[0, 0].set_ylabel("y [cm]")
- # print(worldBoundsX, worldBoundsY)
grid.set(xlim=(-worldBoundsX, worldBoundsX))
grid.set(ylim=(-worldBoundsY, worldBoundsY))
grids.append(grid)
@@ -636,12 +643,12 @@ def evaluate_follow_iid(
}
)
+ plt.rcParams["lines.markersize"] = 1
grid = sns.jointplot(
x="IID [cm]",
y="Follow",
data=follow_iid_data,
linewidth=0,
- s=1,
kind="scatter",
xlim=(0, maxDist),
ylim=(-5, 5),
@@ -682,7 +689,7 @@ def evaluate_all(
predicate: a lambda function, selecting entities
(example: lambda e: e.category == "fish")
"""
- save_folder = Path(save_folder)
+ # save_folder = Path(save_folder)
evaluate_speed(paths, labels, save_folder + "speed.png", predicate)
evaluate_turn(paths, labels, save_folder + "turn.png", predicate)
evaluate_orientation(paths, labels, save_folder + "orientation.png", predicate)